Dario Dario Bressanini Bressanini Qmc in the Apuan Alps III Qmc in the Apuan Alps III ( ( Vallico sotto Vallico sotto ) ) 2007 2007 http://scienze-como.uninsubria.it/ http://scienze-como.uninsubria.it/ bressanini bressanini Universita’ dell’Insubria, Como, Italy Universita’ dell’Insubria, Como, Italy The quest for compact and accurate The quest for compact and accurate trial wave functions trial wave functions
Universita’ dell’Insubria, Como, Italy. The quest for compact and accurate trial wave functions. Dario Bressanini. http://scienze-como.uninsubria.it/ bressanini. Qmc in the Apuan Alps III ( Vallico sotto ) 2007. 30 years of QMC in chemistry. The Early promises?. - PowerPoint PPT Presentation
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Dario Dario BressaniniBressanini
Qmc in the Apuan Alps III Qmc in the Apuan Alps III ((Vallico sottoVallico sotto) ) 2007 2007
The quest for compact and accurateThe quest for compact and accurate trial wave functions trial wave functions
2
30 years of QMC in 30 years of QMC in chemistrychemistry
3
The Early promises?The Early promises?
Solve the Schrödinger equation Solve the Schrödinger equation exactly exactly withoutwithout approximationapproximation (very strong)(very strong)
Solve the Schrödinger equation with Solve the Schrödinger equation with controlled approximationscontrolled approximations, and converge , and converge to the exact solution to the exact solution (strong)(strong)
Solve the Schrödinger equation with Solve the Schrödinger equation with some some approximationapproximation, and do better than other , and do better than other methods methods (weak)(weak)
4
Good for Helium studiesGood for Helium studies
ThousandsThousands of theoretical and experimental of theoretical and experimental paperspapers
)()(ˆ RR nnn EH )()(ˆ RR nnn EH
have been published on Helium, in its various forms:have been published on Helium, in its various forms:
AtomAtom Small ClustersSmall Clusters DropletsDroplets BulkBulk
5
Good for vibrational Good for vibrational problemsproblems
6
For electronic structure?For electronic structure?
Sign ProblemSign Problem
Fixed Nodal error problemFixed Nodal error problem
8
What to do?What to do?
Should we be happy with the “cancellation Should we be happy with the “cancellation of error”, and pursue it?of error”, and pursue it?
If so:If so: Is there the risk, in this case, that QMC Is there the risk, in this case, that QMC
becomes becomes Yet Another Computational ToolYet Another Computational Tool, and , and not particularly efficient nor reliable?not particularly efficient nor reliable?
If not, and pursue If not, and pursue orthodox QMCorthodox QMC (no (no
pseudopotentials, no cancellation of errors, …)pseudopotentials, no cancellation of errors, …) , can we , can we avoid theavoid the curse of curse of TT ? ?
9
The curse of The curse of
QMC currently heavily relies on QMC currently heavily relies on TT(R)(R)
Walter Kohn in its Nobel lecture Walter Kohn in its Nobel lecture (R.M.P. (R.M.P. 7171, 1253 , 1253
(1999))(1999)) tried to “discredit” the wave function as a tried to “discredit” the wave function as a non legitimate conceptnon legitimate concept when when NN (number of (number of electrons) is largeelectrons) is large
1033 ppM N 1033 ppM N
pp = parameters per variable = parameters per variable
MM = total parameters needed = total parameters needed
For For MM=10=1099 and and pp=3=3 NN=6=6
The Exponential WallThe Exponential WallThe Exponential WallThe Exponential Wall
10
The curse of The curse of
Current research focusses onCurrent research focusses on optimizing the energy for moderately large optimizing the energy for moderately large
expansions (good results)expansions (good results) Exploring new trial wave function forms, with a Exploring new trial wave function forms, with a
moderately large number of parameters (good moderately large number of parameters (good results)results)
Is it hopeless to ask for both Is it hopeless to ask for both accurateaccurate and and compactcompact wave functions? wave functions?
12
LiLi2 2 J. Chem. Phys. J. Chem. Phys. 123123, 204109 (2005), 204109 (2005)
-14.9954 E (N.R.L.)E (N.R.L.)
-14.9952(1)
-14.9939(2)
-14.9933(1)
-14.9933(2)
-14.9914(2)
-14.9923(2)
E (hartree) CSF22 g
222 9...43 ggg 22 11 uyux
224 uyux nn 222 211 uuyux
2222 3211 guuyux
Not all CSF are usefulNot all CSF are useful Only 4 csf are needed to build a statistically Only 4 csf are needed to build a statistically
exact nodal surfaceexact nodal surface
(1(1gg22 1 1uu
22 omitted) omitted)
13
A tentative recipeA tentative recipe Use a large Slater basisUse a large Slater basis
But not too largeBut not too large Try to reach HF Try to reach HF nodesnodes convergence convergence
Orbitals from CAS Orbitals from CAS seemseem better than HF, or NO better than HF, or NO Not worth optimizing MOs, if the basis is large Not worth optimizing MOs, if the basis is large
enoughenough Only few configurationsOnly few configurations seem to improve the FN seem to improve the FN
energyenergy Use the right determinants...Use the right determinants...
A wave function with the correct asymptotic A wave function with the correct asymptotic conditions:conditions:
Bressanini and Morosi: JCP Bressanini and Morosi: JCP 119119, 7037 (2003), 7037 (2003)
)()()()()ˆ1(),2,1(112 ee
rgPsrfHPe )()()()()ˆ1(),2,1(112 ee
rgPsrfHPe
22
BasisBasis
In order to build In order to build compactcompact wave functions we used wave functions we used basis functions where basis functions where the cuspthe cusp and the and the asymptotic asymptotic behaviorbehavior are decoupled are decoupled
r
brar
es
1
2
1 r
brar
es
1
2
10 r
are 0 rare
rbre rbre
r
brar
x exp
1
2
2 r
brar
x exp
1
2
2
Use one function per electron plus a simple JastrowUse one function per electron plus a simple Jastrow Can fix the cusps of the orbitals. Can fix the cusps of the orbitals. Very few Very few
Nodal error Nodal error component in HF component in HF wave function wave function coming from coming from incorrect incorrect dissociation?dissociation?
33
GVB for moleculesGVB for molecules
Localized orbitalsLocalized orbitals
34
GVB LiGVB Li22
E (N.R.L.)E (N.R.L.)
-14.9936(1)-14.9936(1)GVB CI 24 det compactGVB CI 24 det compact
-14.9632(1)-14.9632(1)CI 3 det compactCI 3 det compactCI 3 det compactCI 3 det compact
-14.9688(1)-14.9688(1)GVB 8 det compactGVB 8 det compact
-14.9523(2)-14.9523(2)HF 1 det compactHF 1 det compact
CI 5 det large basisCI 5 det large basisCI 5 det large basisCI 5 det large basis
Improvement in the wave functionImprovement in the wave function
but but irrelevantirrelevant on the nodes, on the nodes,
35
Different coordinatesDifferent coordinates
The usual coordinates The usual coordinates might not be the best to might not be the best to describe describe orbitalsorbitals and and wave functionswave functions
In LCAO need to use In LCAO need to use large basislarge basis
For dimers, For dimers, elliptical elliptical confocal coordinatesconfocal coordinates are are more “natural”more “natural”
AB
iBiAi
AB
iBiAi
R
rr
R
rr
AB
iBiAi
AB
iBiAi
R
rr
R
rr
36
Different coordinatesDifferent coordinates
LiLi22 ground state ground state
Compact MOs built using elliptic coordinatesCompact MOs built using elliptic coordinates
expexp
eccs
ep
es
yx 22
)1(2
2
1
221
expexp
eccs
ep
es
yx 22
)1(2
2
1
221
37
LiLi22
E (N.R.L.)E (N.R.L.)
-14.9632(1)-14.9632(1)CI 3 det compactCI 3 det compact
-14.9523(2)-14.9523(2)HF 1 det compactHF 1 det compact -14.9916(1)-14.9916(1)
-14.9931(1)-14.9931(1)
-14.9954
CI 3 det ellipticCI 3 det elliptic -14.9937(1)-14.9937(1)-14.9670(1)-14.9670(1)
HF 1 det ellipticHF 1 det elliptic -14.9543(1)-14.9543(1) -14.9916(1)-14.9916(1)
VMCVMC
Wave functionsWave functions
DMCDMC
Some improvement in the wave functionSome improvement in the wave function
but but negligiblenegligible on the nodes, on the nodes,
38
Different coordinatesDifferent coordinates It It mightmight make a make a
difference even on difference even on nodes for etheronucleinodes for etheronuclei
Consider LiHConsider LiH+3+3 the 2s the 2s state:state:
HF LCAOHF LCAO
HHLiLi
The wave function is The wave function is dominated by the 2s on Lidominated by the 2s on Li
The node (in The node (in redred) is ) is asymmetricalasymmetrical
However the exact node However the exact node must be must be symmetricsymmetric
39
Different coordinatesDifferent coordinates
This is an explicit example of a phenomenon already This is an explicit example of a phenomenon already encountered in other systems, encountered in other systems, the symmetry of the the symmetry of the nodenode is is higherhigher than the than the symmetry of the wave symmetry of the wave functionfunction
The convergence to the exact node, in The convergence to the exact node, in LCAOLCAO, is , is very very slowslow..
Using elliptical coordinates is the right way to Using elliptical coordinates is the right way to proceedproceed
HF LCAOHF LCAO
HHLiLi
Future work will explore if Future work will explore if this effect might be this effect might be important in the important in the construction of many body construction of many body nodesnodes
40
Playing directly with nodes?Playing directly with nodes?
It would be useful to be able to optimize only It would be useful to be able to optimize only those parameters that alter the nodal structurethose parameters that alter the nodal structure
A first “A first “explorationexploration” using a simple test system:” using a simple test system:
HeHe22++
21)1( cc 21)1( cc
The nodes seem to The nodes seem to be smooth and be smooth and ““simplesimple””
Can we “expand” the Can we “expand” the nodes on a basis?nodes on a basis?
41
HeHe22++: “expanding” the node: “expanding” the node
0:)( 1 cNode
BABA rrrr 3311
1:)( 2 cNode
031 zz
Exact
It is a one It is a one parameter parameter
42
““expanding” nodesexpanding” nodes
This was only a kind of “This was only a kind of “proof of conceptproof of concept”” It remains to be seen if it can be applied to It remains to be seen if it can be applied to
larger systemslarger systems Writing “simple” (algebraic?) trial nodes is not Writing “simple” (algebraic?) trial nodes is not
difficult ….difficult …. Waierstrass theoremWaierstrass theorem The goal is to have only The goal is to have only few linearfew linear parameters to parameters to
optimizeoptimize Will it work???????Will it work???????
44
ConclusionsConclusions
The wave function can be improved by The wave function can be improved by incorporating the known analytical incorporating the known analytical structure… structure… with a small number of with a small number of parametersparameters
… … but the nodes do not seem to improve but the nodes do not seem to improve It seems more promising to directly It seems more promising to directly
“manipulate” the nodes.“manipulate” the nodes.
45
A QMC song...A QMC song...
He deals the cards to find the answersHe deals the cards to find the answers
the sacredthe sacred geometry of chancegeometry of chance
thethe hidden lawhidden law of a probable outcomeof a probable outcome
the the numbersnumbers lead a dance lead a dance
Sting: Shape of my heartSting: Shape of my heart
46
JustJust an example an example
Try a different representationTry a different representation Is Is somesome QMC in the momentum QMC in the momentum
representationrepresentation Possible ? And if so, is it:Possible ? And if so, is it: Practical ?Practical ? Useful/Advantageus ?Useful/Advantageus ? Eventually better than Eventually better than plain vanillaplain vanilla QMC ? QMC ? Better suited for some problems/systems ?Better suited for some problems/systems ? Less plagued by the usual problems ?Less plagued by the usual problems ?
47
The other half of Quantum The other half of Quantum mechanicsmechanics
))((ˆ)( rFp ))((ˆ)( rFp The Schrodinger equation in the momentum representationThe Schrodinger equation in the momentum representation
pdpppVpm
pE )()(ˆ)2()()
2( 2/1
2
pdpppVpm
pE )()(ˆ)2()()
2( 2/1
2
SomeSome QMC (GFMC) should be possible, given the iterative form QMC (GFMC) should be possible, given the iterative form
OrOr write the imaginary time propagator in momentum space write the imaginary time propagator in momentum space
48
Better?Better? For coulomb systems:For coulomb systems:
2
2)1(ˆ)(ˆ
jiij pprFpV
2
2)1(ˆ)(ˆ
jiij pprFpV
There are There are NO cuspsNO cusps in momentum space. in momentum space. convergence should be fasterconvergence should be faster
Hydrogenic orbitals are Hydrogenic orbitals are simple rational functionssimple rational functions
222
2/15
1 )(
)8()(
Zp
Zps
222
2/15
1 )(
)8()(
Zp
Zps
54
Avoided nodal crossingAvoided nodal crossing
At a nodal crossing, At a nodal crossing, and and are zero are zero Avoided nodal crossing is the rule, not the Avoided nodal crossing is the rule, not the
exceptionexception Not (Not (yetyet) a proof...) a proof...
0
0
0
0variablesNwithN
eqsN
eq313
.3
.1
variablesNwithNeqsN
eq313
.3
.1
IfIf HFHF has 4 nodeshas 4 nodes HF HF has 2 nodes, with a properhas 2 nodes, with a proper
In the generic case there is no solution to these equationsIn the generic case there is no solution to these equations
55
He atom with noninteracting He atom with noninteracting electronselectrons
Sss 153 Sss 153
56
59
How to How to directlydirectly improve improve nodes?nodes?
““expand” the nodes and optimize the expand” the nodes and optimize the parametersparameters
IFIF the topology is correct, use a coordinate the topology is correct, use a coordinate transformationtransformation
Take a wave function with the correct nodal Take a wave function with the correct nodal topologytopology
)(RR T )(RR T
HF HF
Change the nodes with a coordinate Change the nodes with a coordinate transformation transformation (Linear? Backflow ?) (Linear? Backflow ?) preserving preserving the topologythe topology