Small Molecular Ions
Johanna Brinne Roos
TheoreticalChemistry
School of Biotechnology
Royal Institute ofTechnology
SmallMolecularIons
Licentiatethesis
c
JohannaBrinneRoos,2007 ISBN:978-91-7178-681-4
PrintedbyUniversitetsserviceUS AB,
Stockholm, Sweden, 2007 Typesetin L A T E Xbytheauthor.
InthisthesisIhavetheoreticallystudiedelectronrecombinationprocesseswith
smallmolecularions. Inthesekindofprocessesresonantstatesareinvolved. To
calculatethepotentialenergyforthese statesasafunction ofinternuclear
dis-tance, structurecalculationsandscatteringcalculationshavetobeperformed.
So far I have been studying the ion-pair formation with in electron
recombi-nation with
H
+
3
. The cross section for this process has been calculated using dierent kind of models, both a time dependent quantum mechanical and asemiclassical. I have also studied the direct process of dissociative
recombi-nation of
HF
+
. To calculate the total cross section for this process, wehave
performedwavepacket propagation onthirtyresonant statesandsummed up
theindividual crosssectionsfor thesestates. Thecrosssectionsforboththese
processes have a similar appearance to those measured experimentallyin the
Paper I:
Ion-pair formation in electron recombinationwith
H
+
3
Å. Larson,J.RoosandA. E.Orel
Phil. Trans. R.Soc. A, 364,2999(2006)
Paper II:
Electroncollisionswith
H
+
3
: ion-pair formationJ.B.Roos,Å.LarsonandA.E.Orel
(Manuscript,tobesubmittedto Phys. Rev. A)
Paper III:
Dissociativerecombination of
HF
+
J.B.Roos,Å.LarsonandA.E.Orel
I wouldrstliketothankHansÅgrenforgivenmetheopportunityto
accom-panyÅsaLarson,mysupervisor,whenshemovedtotheTheoreticalChemistry
departmentin 2006.
ÅsaLarsonandAnnOrel,Ireallyenjoyworkingwithyou. Beingamother
oftwoitisextrainspiringtohaveyouasrolemodels.
The thesis is based on three theoretical papers, one published and two in
manuscriptform.
InpaperI,Iwasinvolvedinthestructurecalculations. Iperformeda
diaba-tizationof thecalculatedadiabaticpotentialenergysurfaces and extrapolated
these surfaces to their asymptotic limits. I was also involved in the
calcula-tion of the electronic couplings between these surfaces and the extrapolation
of these. I alsocalculatedthe classicalreactionpath ontheion-pair potential
energysurface. ThispathIthenextractedfromallcalculatedpotentialsurfaces
andtheachievedcurveswereusedforasemiclassicalcalculationoftheion-pair
reactioncrosssection.
InpaperII,wewereusingMCTDHmethodtopropagatethewavepackets.
Thismethodrequiresthatthepotentialsurfacesandcouplingstobeinproduct
form. Ihavedevelopedafunctionthathastherequiredformandworkswellfor
allthediabatic potentialscalculatedearlierandttedthese tothatfunction. I
wasalsoperformingthettingofallelectroniccouplingstoappropriateforms.
Inpaper III,I have been performing allthestructure calculationsand the
scatteringcalculationsforthestatesofsingletsymmetry. I haveperformedthe
diabatizationofthepotentials. Ihavedeterminedtheasymptoticlimitsforthe
systemand extrapolatedallpotentialsandautoionizationwidths using
appro-priate methods. Finally Iperformedwavepacket propagationontheresonant
states of the system and calculated thecross section for the reaction. I have
1 Introduction 1
1.1 Molecularionsin spaceand industry . . . 1
1.2 Electronrecombinationprocesses . . . 2
2 Theoreticaltreatment 5 2.1 TheBorn-Oppenheimerapproximation. . . 5
2.2 Potentialsurfacesandcouplings. . . 6
2.2.1 Structurecalculations . . . 10
2.2.2 Scatteringcalculations . . . 15
2.2.3 Extrapolation. . . 18
2.3 DissociationDynamics . . . 21
2.3.1 Wavepackettreatment . . . 21
2.3.2 TheMCTDHmethod . . . 26
3 Resultsand Discussion 28 3.1 PaperI . . . 28
3.2 PaperII . . . 29
Introduction
1.1 Molecular ions in space and industry
Dissociative recombination and ion-pair formation are key processes in the
physicsofplasmaswheremolecularionsarepresent. Inionstoragerings,plasma
environmentcanberesembled,and these ringshavebeenused successfullyto
study the above mentioned reactions experimentally [1]. We are comparing
ourtheoreticalcalculatedcrosssection forthesereactionswithmeasuredcross
sectionsusingtheion-storagerings.
Lowtemperatureplasmashavemadeasignicantimpactonsocietyduring
thelast half centuryand improvedthe quality oflife. Fluorescentlightisone
example found in many homes today and plasma ion sources are used to
im-plantionsintomaterialssuchassemiconductorchipsforthecomputerindustry.
Manylow-temperatureplasmaapplications involvecomplexreactionsbetween
electronsandahostofatomic,molecular, andionicspecies. Thesespeciesare
found in highly excited states not encountered in non plasma environments.
Studyingthedynamics onthesestatescangiveusefulinformationtoscientists
modelingthese plasmas[2].
Oneofthemostimportantprocessesindeterminingthespeciccomposition
ofanyionosphereisthedissociativerecombinationofmolecularions.
Dissocia-tiverecombinationiseectivelythe onlyelectronlossmechanismin planetary
ionospheres and thus it playsakey role in the daily variationsin ionospheric
electrondensities. Itisalsotheprimarylossmechanismformolecularions,and
isresponsibleforsignicantalterationsinthepopulationsofthevariousatomic
andmolecularspeciespresent. Sir DavidBateswasin 1950thersttosuggest
that the decayof ionization from the ionosphere couldbe explainedby
disso-ciativerecombination. In order to dosohewasusinga set ofdiabatic states,
1.2 Electron recombination processes
Forrelativelowinteractionenergies,thereareingeneralonlyafewdissociation
processesthatoccur.
Indissociativerecombination (DR),the electronis resonantlycaptured by
theion,loosingitsenergyeitherto electronic-ortorovibronicexcitation
exci-tationoftheresultingneutralmolecule. Theformeroftheseprocessesisknown
asthedirect modeofDRandthelatterastheindirectmodeofDR.
In the direct mode of DR, rst proposed by Bardsley [4], the electron is
captured into a double excited state of the molecule, which has a repulsive
potentialcurve,see gure1.1. Inthisstatethemoleculecaneitherautoionize,
re-emitting the electronto the autoionization continuum, orit can dissociate
intofragments. Theprocesscanberepresentedby:
AB
+
+ e
−
→ AB
∗∗
→ A + B.
(1.1)Intheindirectmodeof DR, theelectroniscaptured intooneofthe highly
excited Rydberg state converging to the the initial ion, see gure 1.2. The
excessenergyistransferredintovibrationalorrotationalexcitationoftheionic
core. If there is a second state of the same type asin the direct mode that
crosstheRydbergstate,predissociationis possible. TheRydbergstatecouple
to the resonant state by electronic coupling and the process is completed by
dissociationalongthisstate. Thisprocesscanberepresentedby:
AB
+
+ e
−
→ AB
∗
→ AB
∗∗
→ A + B.
(1.2)InDR, bydenition,neutralfragmentsareformed. Theion-pairformation
process, sometimes referred to as RIP (Resonant Ion-Pair formation), is the
samekindofprocessasDRexceptthationicfragmentsareformed. Therelevant
that, if the electron anity for the
B
fragment in theA
+
+ B
−
channel of
RIP is larger than the dissociation energy of the initial ion, this channel is
openfordissociationatzerointeractionenergy. Thedirect RIPprocesscanbe
representedby:
AB
+
+ e
−
→ AB
∗∗
→ A
+
+ B
−
,
(1.3)andtheindirectprocessofRIPcanberepresentedby:
AB
+
+ e
−
→ AB
∗
→ AB
∗∗
→ A
+
+ B
−
.
(1.4)Inelectronrecombinationwiththemolecular ion
AB
+
, thereis aCoulomb
attractionbetweentheinitialionandtheelectronandthereforethecrosssection
will behigh at low interactionenergies. Itwasshown in 1948 by E.P Wigner
that foraCoulombattractionbetweencollidingparticlesthecrosssection will
decreasewith
1/E
atsmall interactionenergies[5].With increasing interaction energy additional processes can occursuch as
dissociativeexcitation (DE). Thisis aninelastic scatteringprocess wherepart
ofthekineticenergyoftheelectronistransferredtoelectronicexcitationofthe
ion. If thisstateis repulsiveneutralandionicfragmentswillbeformed. This
reactioncanberepresentedby:
AB
+
+ e
−
→ (AB
+
)
∗
+ e
−
→ A + B
+
+ e
−
Figure 1.3: Schematic potentialcurvesfor theRIP process.
D
0
(AB
+
)
is the
Theoretical treatment
Bycalculatingthepotentialenergysurfaceandstudythenucleardynamicson
thatsurface,theoutcomeofchemicalreactionscanbestudied. Ifthepotential
energy surfaces are coupled, dierent outcomes(products) of the reactionare
possible. With wavepacketswecanstudythemotiononcoupledpotential
en-ergysurfacesandwecaninpracticestopthereactionatanytimeandanalyze
how theinitial wavepacket haspropagated, spread outand been distributed
among theelectronic states. Wavepacketsare thus adoortowardsbetter
un-derstanding of quantum mechanics and towards quantum control of reaction
dynamics.
2.1 The Born-Oppenheimer approximation
Ifallnonelectrostaticinteractionsareignored,theHamiltonianforanymolecule
canbewrittenas
H
en
= T
n
+ T
e
+ V
en
+ V
ee
+ V
nn
,
(2.1)where
T
n
is the nuclear kinetic energy,T
e
the electronic kinetic energy,V
en
theelectron-nuclearelectrostaticinteraction,V
ee
theelectron-electronCoulomb repulsionandV
nn
thebarenucleus-nucleusCoulombrepulsion.ThefamousBorn-Oppenheimerapproximation[6],basedontheassumption
that the electronsin general move much faster than themuch heavier nuclei,
separatesthetimeindependentSchrödingerequationintoonenuclear-andone
electronic part.
Theelectronic Schrödingerequationforadiatomicmoleculehastheform
[H
e
− E
i
e
(R)]Φ
i
(R, r) = 0,
(2.2)where
H
e
= T
e
+ V
en
+ V
ee
+ V
nn
.
(2.3)Here, R is the internuclear distance. It appear no longer asavariable in the
equation,insteaditisaparameter. BysolvingtheelectronicSchrödinger
equa-tionatdierentnuclearpositions,aBorn-Oppenheimerpotentialcurve,
E
e
i
(R)
statessincetheseareingeneralwellseparatedfromtheexcitedstates. Forthe
excited states, on the other hand, the potential surfaces are not always well
separatedfromeachotherandthestatesinteractmoreeasily. Inthoseregions,
nucleardynamicswillnotfollowtheBorn-Oppenheimerstatesadiabaticallyand
theBorn-Oppenheimerapproximationbreaksdown.
Theinteractionof thenuclearand electronicmotionsisresponsibleforthe
couplingsbetweendierentneutralelectronicstatesandalsoforautoionization.
Whendescribingprocessesasdissociativerecombinationandion-pairformation,
these couplingsarecrucial.
Thenucleartime-independentScrödingerequation
[T
n
+ E
e
j
(R)]χ
j
(R) = E
j
χ
j
(R)
(2.4)describesthenuclearmotiononstate
j
. Itcanbethevibrationalorrotational motionofboundelectronicstatesandalsodissociationdynamicsondissociativestates. Inthisthesis,thedissociationdynamicshasbeentreatedwithdierent
methods, seesection2.3.
2.2 Potential surfaces and couplings
If we add a small diagonal correction, like a perturbation, to the electronic
energy calculated with the Born-Oppenheimer approximation, that take into
accounttheweakcouplingbetweentheelectronicandnuclearmotions,weend
upwiththeadiabaticpotentialsurface:
E
ad
i
(R) ≡ E
i
e
(R) + hΦ
ad
i
|T
n
|Φ
ad
i
i.
(2.5) The electronic part of the Hamiltonian in this approximation is diagonal, apropertythat makesthesepotentialsfairlyeasytocalculate
hΦ
ad
i
|H
e
|Φ
ad
j
i = E
i
e
(R)δ
ij
.
(2.6)Theodiagonalelementsofthenuclearkineticenergyoperatorinthis
approx-imationdierfromzeroandcausecouplingsbetweentheadiabaticstates
hΦ
ad
i
|T
n
|Φ
ad
j
i 6= 0.
(2.7) Thesearecalledthenon-adiabaticcouplingelements. Itwasshownin1929,byNeumannandWignerthattheadiabaticpotentialenergycurves,foradiatomic
molecule,correspondingtoelectronicstatesofthesamesymmetrycannotcross
[7]. Thisisreferredtoasthenon-crossingrule. Instead,twoadiabaticpotential
energycurvesofthesamesymmetryrepeleachotherwhentheycomeclose. The
pointofclosestapproachiscalledanavoidedcrossingpoint,
R
x
in gure2.1. Thestrongestcontributiontothenon-adiabaticcoupling,givenbyequation(2.7), in the regionof an avoided crossingcomes from arst derivativeradial
couplingterm[8], thatforeach internuclearcoordinatehastheappearance
−
1
µ
hΦ
ad
i
|
d
dR
|Φ
ad
j
i
r
d
dR
.
(2.8)Byincludingnon-adiabaticcouplingtermswewillgetanadiabaticSchrödinger
adiabatic potentialcurvesand dotted lines fordiabatic potential curves.
Φ
ad
and
Φ
d
representtheadiabaticanddiabatic electronicwavefunctions.
dierentialequation ofthat kindismoredicultto solvethanonecontaining
only couplings of potential form. In addition, the numerical evolution of the
couplingscanalsobenontrivial.
Byinsteadusediabaticstatesdenedinsuchawaythattherstderivative
couplingsdisappearwecangetaroundthisproblem. Therstonetousestates
that later would be referredto asdiabatic states wasZener in 1932 [9]. The
adiabaticelectroniceigenfunctionshaveatypicalchangeofcharacterclosetoan
avoidedcrossing. Zenerassumedthatthebasisfunctionsusedwere
approxima-tiveeigenfunctionsoftheelectronicHamiltonianwithoutthischaracteristic,as
showningure2.1. Asthenamediabaticindicates,thesestatesdonotadjust
adiabaticallytochangesoftheinternucleardistance.
However,ashas beenpointedout byC. A. Mead and D. G. Truhlar [10],
everycompletesetofelectronicstatesfullling
hΦ
di
i
|
d
dR
|Φ
di
j
i
r
= 0
(2.9)mustbeindependentofR,butinacalculation,thebasissetusedisbothnite
andincompleteandthenadiabatizationwithinthenumberofstatesconsidered
ispossible.
Thediabaticstateswillbecoupledbytheo-diagonalelementsofthe
elec-tronicHamiltonian:
c
ij
= hΦ
di
i
|H
e
|Φ
di
j
i
r
6= 0.
(2.10)Ifwestartbyassumingthat theadiabaticand diabaticstatesareequalfar
matrixVtothediabaticpotentialmatrixUcanbedoneasfollows[11]
U
= MVM
−
1
,
(2.11) whereM
=
cos[γ(R)]
sin[γ(R)]
− sin[γ(R)] cos[γ(R)]
.
(2.12)TheadiabaticpotentialmatrixVforthetwostateproblem is
V
=
E
ad
1
0
0
E
ad
2
,
(2.13)sotheelementsofthediabaticpotentialmatrixUwillbecome:
U
11
=
E
1
di
(R) = E
1
ad
(R) cos
2
γ(R) + E
2
ad
(R) sin
2
γ(R)
U
22
=
E
1
di
(R) = E
1
ad
(R) sin
2
γ(R) + E
ad
2
(R) cos
2
γ(R)
U
12
=
U
21
= c
12
= c
21
=
1
2
(E
ad
2
(R) − E
1
ad
(R)) sin(2γ(R))
(2.14)Thus, thediagonal elements of thediabatic potential matrix are the diabatic
potentials and the o-diagonal elements are the electronic couplings. From
the equationabovewecansee that, at the curvecrossing point
R
x
, the angleγ = π/2
andtheelectronic couplingwillbegivenbyhalf thedistancebetween theadiabaticstates,arelationthatisusedintheLandau-Zenermodel,appliedin paperI.
ForallR,theangle
γ(R)
canbeobtainedfromthecongurationinteraction (CI)coecients. Ifweonlyhavetwointeractingadiabaticstatesdominatedbytwo congurations, we havethe following relation between the adiabatic and
diabaticstates:
Φ
ad
1
Φ
ad
2
=
C
11
C
12
C
21
C
22
Φ
di
1
Φ
di
2
.
(2.15)Sincethe adiabaticanddiabaticwavefunctions areorthonormaltheabove
re-lationcanalsobewritten as
Φ
ad
1
Φ
ad
2
=
cos[γ(R)] − sin[γ(R)]
sin[γ(R)]
cos[γ(R)]
Φ
di
1
Φ
di
2
,
(2.16)andwecanseethattheangle
γ(R)
isgivenbytan[γ(R)] =
C
21
(R)
C
11
(R)
.
(2.17)
The above expressions are developed for diatomic molecules and canbe used
if the potential curves are well separated. In paper I and II, we have also
usedtheaboverelationstoestimatethecouplingbetweentheion-pairpotential
energysurfaceinelectronrecombinationwith
H
+
3
andthelowerRydbergstates convergingtoH
+
3
. Inthis paperweused the radialJacobi coordinatesr
andFigure 2.2: Eectivequantum numbersforRydbergstatesof
1
Σ
+
symmetry
convergingtothegroundstateof
HF
+
one,
r
, xed. The electronic coupling between the states,c
12
, wascalculated withequation(2.14).In dissociative recombination, the electron is captured by the ion into a
dissociativemetastable statethat sometimescrossthewholemanifoldof
Ry-dbergstatesconvergingtotheinitialion. Itismetastable inasensethatitis
associatedwith alifetime andhenceawidth
Γ
. ForthehigherRydbergstates thatareverycloseinenergy,itisoftennotpossibletoassumethatthecrossingstatesonlycoupletwo-by-two.Wecanhoweverestimatetheelectroniccoupling
between theresonant andRydberg states by scaling the autoionizationwidth
Γ(R)
[12]hΦ
ryd
(R, r)|H
e
|Φ
d
(R, r)i
r
=
r
Γ(R)
2π
1
(n
∗
(R))
3/2
.
(2.18) Here,n
∗
(R)
istheeectivequantumnumberoftheRydbergstatethat, accord-ingtoMulliken'sformula[13]isgivenbyn
∗
(R) = n − µ(R) =
1
p2(E
ion
(R) − E
ryd
(R))
.
(2.19)In this expression,
n
andµ
are the principal quantum number and quantum defect of the Rydberg state respectively,E
ion
(R)
is the initial ion potential energyandE
ryd
(R)
istheRydbergstatepotential. Ingure(2.2),weshowthe eectivequantum numbersofthediabatiziedRydbergstatesofHF.Ifthekineticenergy ishigh at acurvecrossingand wehavealarge
adiabaticrepresentationwouldbetterimitatethetrueevolvementofthesystem.
Inelectronrecombinationprocesses,wenormallyhavetherssetofconditions,
andwethereforechose adiabaticrepresentation.
2.2.1 Structure calculations
Conguration interaction Calculation
The electronic structure calculation starts with a Hartree-Fock (HF)
calcula-tion,fromwhichmolecularorbitals(MOs)andagroundstatesolutionthatnot
accuratelyinclude correlationisobtained.
The conguration interaction method (CI) is the simplest way to include
electroncorrelationandtoimprovetheHartree-Focksolution. Thisisamethod
wellsuited forcalculatingexcitedstatesofsmallmolecules.
ThewavefunctionisconstructedasalinearcombinationofSlater
determi-nantsorcongurationstatefunctions(CSF)
Ψ
CI
=
N
SCF
X
m=1
C
m
Ψ
m
.
(2.20)CSFs are createdby distributingthe electrons in the MOs obtained from the
Hartree-Focksolution
Ψ
CI
= C
o
|HF i +
occ.
X
i
virt.
X
r
C
i
r
Ψ
r
i
+
occ.
X
i<j
virt.
X
r<s
C
ij
rs
Ψ
rs
ij
+ . . . .
(2.21)where
occ.
standsforoccupiedorbitalsandvirt.
forvirtualorbitals. The vari-ationalprincipleis thenusedforsolvingtheelectronicSchrödingerequationE
var
=
R Ψ
∗
H
e
Ψdτ
R Ψ
∗
Ψdτ
≥ E
0
true
(2.22)Foralineartrialwavefunction,thevariationalprincipleleadstosolvingthe
secularequationfortheCIcoecientsordiagonalizingtheCImatrix.
ForafullCI(FCI),thecompletesetofdeterminantsgeneratedby
distribut-ing the electrons among all orbitals is included in the expansion(2.21). The
numberof Slater determinants increaseveryrapidly with thenumber of
elec-trons and with the numberof orbitals. A FCI expansionis therefore suitable
onlyforthesmallestelectronicsystems. InpaperIandII,afullCIcalculation
wasused to determinethe adiabaticpotentialsof
H
3
situated belowthe ionic groundstateofH
+
3
.Forlargerelectronicsystem,theFCIexpansionhastobetruncated. When
designing smaller congurationspaces, it is important to distinguish between
staticand dynamicalcorrelation. Static correlationis treatedbyretaining the
dominant congurationsof theFCI expansionaswell asthose that arenearly
degeneratewiththedominantcongurations. Thesecongurationsarereferred
to as referencecongurations of theCI wave function, andthey span a
refer-ence space. Dynamical correlation is treated by adding to the wavefunction
allcongurationsthatmaybecomeimportantintheregionofaenergysurface
neededtodescribeacertainreaction. TheMRCIwavefunctionisgeneratedby
addingtoallcongurationsinthisreferencespaceallexcitationsuptoagiven
level from each reference conguration. In paper III, the MRCI method was
usedyodeterminetheHFpotentialsbelowthegroundstateof
HF
+
.
Validation
Below the ion state potential, structure calculationscan be used to calculate
theessentialpotentialenergysurfaces ofthesystemunderstudy.
Thisisnotpossibleforthestatesabovetheion,sincetheseresonantstates
are associatedwith awidth, related to thelifetime of these states. They are
referredto asresonantstates sincetheyshowupasa sharpvariation, a
reso-nance, in thecross sectionfor elasticscattering betweenthe electronand ion.
Withastructurecalculation,wecannotbesurethatwehavereceivedtheexact
positionoftheresonantstate. Thepositionwillbewithinthewidth,andifthe
widthisnarrow,astructurecalculationwillnotgiveasignicanterror. Onthe
otherhand,weneedtocalculate thewidthtondoutifwemakeasignicant
error,andforthiswehavetorunscatteringcalculations.
Scatteringcalculationsarefarmoretimeconsumingthanstructure
calcula-tions. Tond theresonance,weneedto perform anenergyscan andsincethe
resonancescanbeverynarrowweneedtouseaverynegridtomakesurewe
donotmissaresonance. Theresonancealsohastobewellresolved,andforthis
at least ten energy points aroundthe resonanceare needed. In paper III, for
HF,eachenergypointtookapproximatelytwohourstocalculate. Tominimize
the time put into the calculation, it is therefore a good idea to start with a
structurecalculationtogetanapproximate position oftheresonance. Forthis
purpose,structure calculationstodeterminetheresonantstatesarejustied.
Choice ofbasis set and wave functionin DR and RIP
Indissociativerecombinationandion-pairformation,therearesometimesmany
dissociativestatesandRydbergstatesofdierentsymmetriesthatareimportant
for the process. In paper III, where dissociative recombination with
HF
+
is
studied, for example,allof the30calculatedresonantstatescontributeto the
totalcrosssection. Itcanalsobethecasethattheresonantstatecoupletothe
whole manifold of Rydberg states of the same symmetry. To represent all of
these states correctly, largebasis sets haveto be used. ForHF, upto 15000
congurations had to be included. For somesystems morethan 20 adiabatic
roots have to be calculated for each symmetry and internuclear distance to
obtainallneutralstatesneededbothbelowand abovethe groundstateofthe
ion.
Continuumstates
Abovethegroundstateoftheion,thecalculatedstatescanbeeitherthe
reso-nantstatesweare interestedin orit canbecontinuum states,i.e. stateswith
the same character asthe ground state of the ion but with anextra electron
diabatizationoftheadiabaticstatesareperformed,thesestatesareremoved.
The choice ofsymmetry
The symmetry of molecules is described in terms of elements and operations.
Elements aregeometric entities such asaxes, planes and points in space used
to dene symmetryoperations. Operationsinvolvethe spatial re-arrangement
of atoms in amoleculebyrotationabout anaxis,
C
n
, by reectionthrougha plane,σ
or by inversionthroughapoint,i
. A rotation,reectionorinversion operation will becalled a symmetryoperationif, and onlyif, the new spatialarrangementoftheatomsinthemoleculeisindistinguishablefromtheoriginal
arrangement.
Thepointgroup,orsymmetrygroup,isthename ofaacollectionof
sym-metryelementspossessedby amolecule. Eachcommon collectionof elements
isrepresentedbyasimplesymbol,calledtheSchöniesnotation,indicatingthe
typeofreectionsymmetryand theorderoftheprincipalrotationalaxis.
The
C
2v
notation is indicating that it exist oneC
2
axis and two vertical planescontainingthisaxis. Byexcludingbendingof themolecule,wecouldinpaperIII carryoutourcalculationsonthe
H
3
moleculeinC
2v
symmetry[14]. ThesymmetrygroupofHFisC
∞
v
, butin paperIIIwecarryoutcalculations onthesysteminC
2v
symmetrysincewefreelycanchosethenumberofvertical reectionplanescontainingtheprincipalaxis.In
C
2v
symmetry,wecalculatestatesofA
1
,A
2
,B
1
andB
2
symmetries. The stateswegetinB
1
andB
2
symmetriesaredegenerateforHF.Ifwewouldlike to go fromC
∞
v
symmetry toC
2v
symmetry, that is transform into states ofΣ
+
,
Σ
−
∆
andΠ
symmetries,thiscanbedoneasfollows:1. Removestatesthat occurinboth
A
1
andA
2
fromA
1
. -Thestatesthataretakenawayare∆
states.-Theremainingstatesare
Σ
+
states.
2. Removethe
∆
statesfrom theA
2
states. -TheremainingstatesareofΣ
−
symmetry.
3. Degeneratestates
B
1
andB
2
givedegeneratestatesofΠ
x
andΠ
y
symme-tries.Diabatization
Thenon-crossing ruletellsus thattwoadiabaticstatesof thesamesymmetry
cannot cross [7]. Ingure 2.3 the resultingHF states of acalculation in
1
A
1
symmetry are shown. Before adiabatizationis performedon these states, we
have to separate the states of
1
Σ
+
and
1
∆
symmetry and remove continuum
states. That yieldsadiabaticstatesof
1
Σ
+
symmetryand
1
∆
symmetry.
A trivial diabatization can be done by at each internuclear distance,
R
i
, identify which root to connect with a root calculated at the greatest lowerinternucleardistance
R
i−1
. If wehaveclearavoidedcrossingsin theadiabatic states,thereceiveddiabaticstateswillhavesomeofthischaracteristicin themFigure 2.3: HFpotentialsof
1
A
1
symmetry. Resultsfrom MRCIcalculationshavean avoided crossing, equations(2.17) and (2.15)canbe usedif itcanbe
assumedthat thepotentialscoupletwo-by-two.
Adiabatizationcanbedonebytrackingthedominantcongurationswhen
theinternucleardistance
R
isvaried. TheRydbergstatesallhavethesame con-gurationasthegroundstateoftheionplusanextraelectroninaouterorbital.Theresonantstateshavecongurationsthat diersfrom thoseoftheRydberg
states. InthiswaywecanseparatetheresonantstatesfromtheRydbergstates.
If we haveindications of avoided crossings betweenthe resonant statesat
someinternucleardistance,
R
,wecanatapointwherethestatesarewell sepa-ratedpickthemostprobablecongurationfor thelowest stateand diabatizisethestatesbytrackingthat congurationtowardssmallerorlargerinternuclear
distancesusingthegeometrydependence oftheCIcoecients. Itcanhowever
bethe case that other stateshave thesame congurationin them too, which
canmakethisproceduredicult. InpaperIII,wefoundthattheconguration
associatedwith the ion-pair in HF was the
(1σ)
2
(2σ)
2
(3σ)
1
(4σ)
1
(1π)
4
cong-uration. This conguration was howeveralso present in other diabatic states
of the same symmetry, see gure (2.4). This gure shows in what states the
(1σ)
2
(2σ)
2
(3σ)
1
(4σ)
1
(1π)
4
congurationappear. The square of the CI
coe-cientsforthiscongurationintheion-pairstatearealsoshown.
Ifwewouldonlyhaveonediabatic statecontainingacertainconguration,
wecouldinsteadofusing equation(2.17)makethediabatic statesmootherby
coecientsforthecongurationasaweight
V (R) =
P
i
C
i
2
(R)V
i
(R)
P
i
c
2
i
(R)
.
(2.23)This method wasused tocreate thediabatic potentialsurfacefor theion-pair
statein
H
3
belowtheionsurfaceinpaperI.For HF(paperIII),weonlyused thetrivialmethodto diabatizatethe
adi-abaticstates. Theresultfor
1
Σ
symmetryisshownin gure2.5. Here wecan
seethatsomecharacteristicforanavoidedcrossingstillremainsinsomeofthe
diabaticstates.
Inafuture study, we areplanningperformamorecareful diabatizationof
theHF potentialsandcalculate the electroniccouplings betweentheresonant
Figure 2.5: Diabatic potentials of
1
Σ
+
symmetry. Diabatization of results
fromMRCIcalculations.
2.2.2 Scattering calculations
Theresonantstatesabovethegroundstateoftheionisassociatedwithawidth,
related to the lifetime of the state. Due to this, these statescannot be
stud-ied usingconventionalstructurecalculations. Weare calculatingthe potential
energy,
V (R)
, andtheautoionizationwidth,Γ(R)
, withscatteringcalculation, using aComplex-Kohnvariational method [15, 16]. Inthese calculations, theelastic scattering process between an incoming electron and the ion target is
studied, see gure2.6. At the resonant energy
E
res
(R)
, the ion canbe tem-porarilycapturedintoanearlyboundstate,which willcauseasharpvariationofthecrosssection,aresonance.
A boundary condition of the of the scattered electronic wave function is
givenby
lim
r→∞
Ψ
k
(r) = e
ik·r
+
e
ikr
r
f (Θ, Φ),
(2.24)where
Θ
andΦ
arethepolarandazimuthalanglesofscatteringrelativetothe directionof incident,ˆ
k
andf (Θ, Φ)
isthescatteringamplitude. Thedierentialcrosssectionforthereactionisgivenbydσ
dΩ
= |f(Θ, Φ)|
2
,
(2.25)
inte-gratingthedierentialcrosssectionoversolidangles:
σ
e
=
Z
dΩ|f(Θ, Φ)|
2
.
(2.26)
Fora sphericalsymmetricpotential, therewill be no
Φ
dependence in the scatteringamplitudeanditispossibleto expandit inthecompletesetofLeg-endrepolynomials,
P
l
(cos Θ)
:f (Θ) =
i
2k
∞
X
l=0
(2l + 1)(1 − S
l
)P
l
(cos Θ),
(2.27)where
S
l
is thescattering,orS
, matrix,which inthecaseofcentral eld scat-teringonlyisaone-by-onematrix,orasinglefunctionofk
,andl
istheangular momentum [17].Thetotalelasticcrosssectionwill become
σ
e
=
π
k
2
∞
X
l=0
(2l + 1)|1 − S
l
|
2
,
(2.28)wherethequantity
1 − S
l
= T
l
isthetransition, orT
,matrix.Thetermresonanceiswidelyusedintheliterature,butunfortunatelywith
dierentmeanings. Itisconvenienttothinkofaresonanceasapoleinthe
S
(orT
)matrix. Apole intheS
(orT
)matrixcorrespondtoapointinthecomplexp
-plane(p = ¯
hk
). Apointinthep
-planecancorrespondtothreedierentkind ofstates:boundstate,
p = iκ
, withκ > 0
virtualstate,p = −iγ
, withγ > 0
resonancestate,p = ±β − iγ
, withβ, γ > 0
When the electronis scattered towardsthe molecular ion, the interaction
potential is non-spherical and the
S
(orT
) matrix will have bothl
andm
indicies.Aresonantstatewillthusshowupasasharpvariationofthecrosssectionat
k
res
accordingtoequation(2.28). Theywillhowevershowupmoreclearlywhen theeigenphasesumisstudied. TheeigenphasesumisrelatedtotheeigenvaluesFigure 2.7: Resonancesin DRwith
HF
+
:
1
A
1
symmetryat internuclear dis-tanceR = 1.2 a
0
andR = 1.3 a
0
(
e
iδ
lm
) of the
S
matrix. As a funct function of energy,E = ¯
h
2
k
2
/2m
, the
eigenphasesumisgivenby:
δ(E) =
X
l,m
δ
l,m
(R, E),
(2.29)where
R
isthexedinternucleardistanceatwhichthecalculationisperformed. At everyenergy where there is a resonance,E
res
, this eigenphase sumjumps suddenlywithπ
. Thiscanbeseeningure2.7. Thisgurealsocleariesan ad-ditionaladvantageofthescatteringcalculations,bytheshapeoftheresonancesit is possible to determine if there exist a crossing between diabatic resonant
statesornot. For
R = 1.2 a
0
wehaveaverynarrowresonanceatE = 0.1375 H
, thatisnotresolvedatR = 1.3 a
0
. Thisresonanceisof1
∆
symmetry,whilethe
otherare of
1
Σ
symmetry. Inthecalculationcarried outin
1
A
2
symmetrythe gridwas neenoughto resolvethisresonance.The
N +1
electronwavefunctionforthesystemshouldsatisfythetime inde-pendentSchrödingerequation. ThetimeindependentSchrödingercanhowevernotbesolvedforsuchasystemsoinsteadweset upatrialwavefunctionwith
unknowncoecients. ThesecoecientsaredeterminedwiththecomplexKohn
variationalmethodand a
T
matrixisobtained. FromthisT
matrix,thecross sectionandeigenphasesumcanbecalculated.Tobeableto mergetheresultsof thescattering calculationsand structure
calculationsweare performingboth calculationswith thesame program. We
Wignerformwegetboththeposition
E
res
(R)
andtheautoionizationwidthΓ
(R)
oftheresonanceatinternucleardistanceR
δ(E)
= δ
res
(E) + δ
bg
(E)
= tan
−
1
Γ
2(E − E
R
)
+ a + bE + cE
2
.
(2.30)Here,
δ
res
isthecontributionfromtheresonancetotheeigenphasesumandδ
bg
isthebackgroundcontribution.Toobtainthepotentialenergyoftheresonantstate,theionicpotentialhas
tobeaddedtotheresonanceenergy
V
res
(R) = V
ion
(R) + E
res
(R).
(2.31)2.2.3 Extrapolation
Scattering calculation are as mention earlier far more time consuming than
structure calculations. First the resonance must be localized, which can be
tricky if it is narrow, then the resonancemust bewell resolved, so the
Breit-Wignerttingwillnotyieldanincorrectresult. Tottheresonance,atleastten
energy pointsare neededto becalculatedclose aroundtheresonance,and for
HF,eachveenergypointstookapproximatelythesamecomputationaleortas
calculating25adiabaticroots. Inaddition,theBreit-Wignerttingtakestime.
Structurecalculationsarehowevertimeconsumingtoo,anddependingonwhich
basissetandnumberofcongurationsthathavebeenusedinthecalculationthe
potentialsurfacecalculatedwillbemoreaccurateatsomeinternucleardistances
than at another. Careful extrapolation is thereforeneeded, bothof potential
surfaces,electronic couplingsandautoionizationwidths.
Towardssmaller internuclear distances
Manypotentialcurvesofdiatomicmolecules canbewelldescribedbyaMorse
potential:
V (R) = D
e
[1 − e
−
α(R−R
e
)
]
2
+ T,
(2.32)where
R
istheinternucleardistance,D
e
isthedissociationenergy,R
e
the equi-libriumbondlength,α
isrelatedtothebondforceconstantandT
thepotential minimumrelativetoareferenceenergylevel,suchasthepotentialminimumofthegroundstateof thesystem. AMorsepotentialisoftenagood
approxima-tion aroundthe potentialminimum and sometimes also at largerinternuclear
distancesbut doesnotbehaveproperlyatsmallinternucleardistances.
AMorse-typepotentialthat behavesbetter atsmall internuclear distances
is
V (R) = D
e
1 −
e
αR
e
− 1
e
αR
− 1
2
+ T.
(2.33)This potentialwas usedwhen the potentaial curvesfor HFwere extrapolated
towards smaller internuclear distances in paper III. This function was
how-everusedonly atinternuclear distanceswhere theFranck-Condonoverlapwas
insignicant. That is, inside the region where the electron capture and the
AtomicStates MolecularStates
S
g
+ S
g
orS
u
+ S
u
Σ
+
S
g
+ S
u
Σ
−
S
g
+ P
g
orS
u
+ P
u
Σ
−
, Π
S
g
+ P
u
orS
u
+ P
g
Σ
+
, Π
S
g
+ D
g
orS
u
+ D
u
Σ
+
, Π, ∆
S
g
+ D
u
orS
u
+ D
g
Σ
−
, Π, ∆
S
g
+ F
g
orS
u
+ F
u
Σ
−
, Π, ∆, Φ
S
g
+ F
u
orS
u
+ F
g
Σ
+
, Π, ∆, Φ
P
g
+ P
g
orP
u
+ P
u
Σ
+
(2), Σ
−
, Φ(2), ∆
P
g
+ P
u
Σ
+
, Σ
−
(2), Φ(2), ∆
P
g
+ D
g
orP
u
+ D
u
Σ
+
, Σ
−
(2), Φ(3), ∆(2), Φ
P
g
+ D
u
orP
u
+ D
g
Σ
+
(2), Σ
−
, Φ(3), ∆(2), Φ
P
g
+ F
g
orP
u
+ F
u
Σ
+
(2), Σ
−
, Φ(3), ∆(3), Φ(2), Γ
P
g
+ F
u
orP
u
+ F
g
Σ
+
, Σ
−
(2), Φ(3), ∆(3), Φ(2), Γ
D
g
+ D
g
orD
u
+ D
u
Σ
+
(3), Σ
−
(2), Φ(4), ∆(3), Φ(2), Γ
D
g
+ D
u
Σ
+
(2), Σ
−
(3), Φ(4), ∆(3), Φ(2), Γ
D
g
+ F
g
orD
u
+ F
u
Σ
+
(2), Σ
−
(3), Φ(5), ∆(4), Φ(3), Γ(2), H
D
g
+ F
u
orD
u
+ F
g
Σ
+
(3), Σ
−
(2), Φ(5), ∆(4), Φ(3), Γ(2), H
Towardsasymptoticlimits
Theasymptoticformofanion-pairstateisnotatinin theasymptoticregion
dueto theCoulombattractionbetweentheion-pair. Instead,thepotentialfor
suchastatehasthefollowingform asymptotically:
V
res
(R) = V
f inal
−
1
R
−
α
2R
4
,
(2.34)where
V
f inal
is the asymptotic energy limit andα
is the polarizability. For HF,weassumedinpaperIIIthat theion-pairpotentialobtainthisasymptoticform at
R ≥ 20 a
0
. We have structure calculationsout toR = 9 a
0
. Spline interpolationisusedtoconnectthetworegions.Ifwehaveadiatomicmolecule,theasymptoticenergylimitsforthe
poten-tial energyare obtainedfrom spectroscopicdata forthe separatedatoms. By
examining thecalculatedpotentialcurvesand using theWigner-Witmer rules
(seetable2.1)[18],wecandeterminewhichasymptoticleveleachresonantstate
goesto. TheWigner-Witmerrulesgiveusthenumberofelectronicstates,ofa
givensymmetry,thatisassociatedwithaspecicasymptoticlimit.
Inaddition, themultiplicityofthemolecularstateshavetobedetermined.
Therelationbetweenthemultiplicitiesoftheatomicstatesandthemultiplicities
forthemolecularstatesaregivenin table2.2.
Finally,theasymptoticenergylevelshavetobeshiftedrelativetosome
refer-encepointinthecalculatedpotentialsandinterpolationbetweenthecalculated
AtomicStates MolecularStates
Singlet+singlet Singlet
Singlet+doublet Doublet
Singlet+triplet Triplet
Doublet+doublet Singlet,triplet
Doublet+triplet Doublet,quartet
Doublet+quartet Triplet,quintet
Triplet+triplet Singlet,triplet,quintet
Triplet+quartet Doublet,quartet,sextet
Quartet+quartet Singlet,triplet,quintet,septet
method. A suitablemethod isnotalwaysaspline,sincethedistance between
thecalculatedpotentialsandtheasymptoticlimitcanbelargeandasplinecan
giveunphysicaloscillatingbehaviourintheintermediateregion.
The ion-pairsurface in
H
3
Fortheion-pairsurfacein
H
3
,calculatedinpaperI,wehavealongthereaction coordinate,z
,theasymptoticlimitH
+
2
+ H
−
. Forlargez,equation(2.34)give
ustheasymptoticformoftheion-pairsurface
V (r, z) −→ E(r) −
1
z
−
2z
α
4
.
(2.35)Here
E(r)
is thepotentialcurveofH
+
2
andα
thepolarizibility ofH
−
. When
the other coordinate,
r
, is increased, we will have the three body breakup,H
+
+ H
−
+ H
, asymptotically. Inthis case,wehaveaCoulombattraction be-tweenH
+
and
H
−
andweassumethattheion-pairpotentialhasthefollowing
form:
V (r, z) −→ E
0
−
1
√
r
2
+ z
2
−
α
2(r
2
+ z
2
)
2
,
(2.36) whereE
0
is thedissociationenergy ofH
+
2
. Ifweinvestigatethe limitwherer
goesto zero,H
+
+ H
−
goesto theunited atom limit
2
He
+
, and wewill have
as a function of
z
, the potential curve for2
HeH
with the asymptotic limit
2
He
+
+ H
−
. With calculatedpotential energypointsaround the minima and
withtheabovelimitsitispossibletoconstructthepotentialion-pairsurfaceof
H
+
3
withasuitableextrapolationmethod.Wecould not usea twodimensional spline to connectthe calculated data
withtheasymptoticlimitsforthesamereasonsasdiscussedearlier. Insteadwe
tted thecalculatedpointsandlimits to atwodimensional function of Morse
character. Theion-pairpotentialcurvein HeHhad been calculatedearlierby
Larsonand Orel [19]. And the potential curvefor
H
+
V (R) =
n
X
j=0
V
j
e
−
jα(R−R
e
)
,
(2.37)where
n = 2
, withthecoecientsV
0
= T + D
e
,V
1
= −2D
e
andV
2
= D
e
. By including terms up ton = 5
it was possible to t the above sum to theH
+
2
potential. If termsup to
n = 9
wasincludedthe function gotexible enough forttingtheion-pairpotentialcurveofHeH.AproductofthesetwofunctionsV (r)
andV (z)
canbewrittenasV (r, z) =
n
i
X
i=0
n
j
X
j=0
V
ij
e
−
iα
r
(r−r
e
)−jα
z
(z−z
e
)
,
(2.38)where
n
i
= 5
andn
j
= 9
. The above function has 49 coecients that need to be determined by tting. In paper I,weused this function to extrapolatethe abinitio calculated ion-pairto its asymptoticlimits. Figure 2.8 showthe
ion-pairsurfacein twodimensions. Also,theclassicalpathwithstartingpoint
at theequilateralgeometryof
H
3
isdisplayed. This pathwascalculatedusing Hamilton'sequationsofmotion.InpaperII,thesurfacewasused directly,sinceithastherequiredproduct
form fortheMCTDHmethod usedforpropagatingwavepacketsinthis paper
[see section2.3.2and equation(2.61)]. Theother, forion-pairformation,
rele-vantpotentialsurfaces in paperII werealsotted successfullyto thefunction
given by equation (2.38). For the same reason, the electronic couplings, and
autoionizationwidthswerettedtoasumofGaussianfunction products:
c(r, z) =
m
i
X
i=0
m
j
X
j=0
c
ij
e
−
α
r
(r−r
i
)
2
−
α
z
(z−z
i
)
2
,
(2.39)where3to6Gaussianfunctionsofeachdimensionwereincludeddependingon
theshapeofthecouplings.
2.3 Dissociation Dynamics
Whenthepotentialsurfacesandcouplingshavebeencalculated,thenuclear
dy-namicscanbestudied. Thisisdonebysolvingthetime-dependentSchrödinger
equationforthesystem. Thefollowingchapterwillcontaininformationonhow
the electron capture, propagation of wave packets and autoionization from a
resonantstatecanbedescribed. Information onhowthe crosssectionfor the
reactioncan beobtainedwillalsobegiven.
2.3.1 Wave packet treatment
The wave packet
The solutionto the Schrödinger equation for afree particle is given by plane
waves. Accordingto theprincipleof linearsuperposition,alinearsuperosition
Figure 2.8: Theion-pairsurfacein electronrecombinationwith
H
+
3
. Theline starting at the equilateral geometry ofH
+
3
indicates the classicalpath on the potentialsurface.asuperpositionin which thewavefunction hasanappreciablemagnitudeonly
overarelativelysmallinterval(areafortwodimensions)andisfallingorapidly
outsidethisinterval[21]. Inonedimension,awavepacket isoftheform
Ψ(R, t) =
√
1
2π
Z
Φ(k)e
i(kR−ω(k)t)
dk,
(2.40)where
k
is the wave number, andω(k)
the angular frequency. If the angular frequencyisexpandasaTaylorseriesaboutthecentralfrequencyk
0
,therst termsintheserieswillbeω(k) = ω
0
+
dω
dk
k
0
(k − k
0
) +
1
2
d
2
ω
dk
2
k
0
(k − k
0
)
2
+ ... .
(2.41)The rstterm is just thecentral frequency, the coecientin thesecond term
is the group velocity, the third term will give us the dispersion of the group
velocity,thespreadingofthewavepacket,andhighertermswillgiveushigher
orderdispersions.
With wavepacket methods,it ispossibletostudy thestatedistributionat
anytimeduring thepropagation andthereforetheygiveenormousinsightinto
the mechanismof the reaction. Thewavepackets, thepotentials,the
autoion-ization widthsand thecouplingsarein these methodsdened ona
R
-grid. In onedimension,thegridisgivenbyWhenthe electronrecombines withthe ion, wavepackets areinitiated onthe
resonantstates. If theelectronic capture is fastcompared with themotion of
the nuclei, an assumptionthat normallyholds, theinitial wave packet on the
resonantstate
i
isgivenbyΨ
i
(t = 0, R) =
r
Γ
i
(R)
2π
χ
v=0
(R),
(2.43)where
χ
v=0
(R)
isthegroundvibrationalwavefunctionfortheion,andΓ
i
(R)
is theautoionizationwidth ofresonantstatei
[22]. AccordingtoFermi'sGolden rule,thepΓ
i
/2π
factoristheelectroniccouplingelementbetweentheresonant stateandtheautoionizationcontinuum.Thevibrational wave fuction forthe ion canbedetermined using a
nite-dierencemethod[23],wherethesecondorderderivativeisapproximatedby
d
2
dR
2
χ(R
i
) ≈
1
(∆R)
2
[χ(R
i+1
− 2χ(R
i
) + χ(R
i−1
)],
(2.44)whichwilltransformthetimeindependentSchrödingerequationintoasolvable
tridiagonalmatrixeigenvalueequation.
Autoionization
Whenthemoleculebeginits fragmentation, theelectroncanbere-emitted,or
autoionize. Thiscanhappenaslongas
R
issmallerthanthecrossingpoint,R
x
, betweentheresonantstateandtheion. Atsucientlyhightotalenergywecanassumethatwehaveclosurefortheenergeticallyopenvibrationallevels. That
is,themoleculecanautoionizeintoallenergeticallyopenvibrationallevels. In
this case wecan apply the boomerang model [24, 25], where autoionization
from theresonantstates isincluded byletting thepotentialsbecomecomplex
abovetheionpotential
V
i
(R) = E
i
(R) − i
Γ
i
(R)
2
.
(2.45)If the total energy is not suciently high, which is typically the case if
the resonantstatecross theion closeto its minimum, thecontribution to the
Hamiltonian ofthe autoionizationhaveto becalculatedfor eachenergy point
usinganon-localoperator[4,26]
OΨ = −iπ
r
Γ(R)
2π
X
v
j
hχ
v
j
(R)|
r
Γ(R)
2π
|Ψ(R)i
R
χ
v
j
(R),
(2.46)wherethesumisoverallenergeticallyopenvibrationallevels.
Propagation
The wave packets are propagated by solving the time-dependent Schrödinger
equation. Inonedimensionwehave
i
∂
Ψ(R, t) =
Ψ
1
(R, t)
Ψ
2
(R, t)
. . .
(2.48)andtheHamiltonianforcoupledpotentialsis givenby
H =
T
n
+ V
1
(R)
c
12
(R)
· · ·
c
21
(R)
T
n
+ V
2
(R) · · ·
. . . . . . . . .
.
(2.49)If the potentials
V
i
(R)
in theHamiltonian are diabatic,c
ij
are theelectronic couplings,givenbyequation(2.11)andc
ij
= c
ji
. Ifthepotentialsareadiabatic, thecouplingelementsaredominatedbytherstderivativeradialcouplingterm,equation(2.7),and
c
ij
= −c
ji
.Thepotentials,theautoionization,thecouplingsandtheinitialwavepacket
arerepresentedbytheirvaluesonthechosengrid. Thekineticenergyoperator
in the Hamiltonian isapproximatedusing thenite dierence method, likein
equation (2.44). The solution to the time dependent Schrödinger equation is
givenby
Ψ(R, t) = e
−
i(t−t
0
)H
Ψ(R, t
0
)
(2.50)where
exp(−i(t − t
0
)H) = exp(−i∆tH)
istheevolutionoperator. Propagation methodsand crosssectionThe propagation of thewavepacket can beperformed with dierent
numeri-cal methods, suchastheCrank-Nicholsonmethod, the SplitOperator Fourier
Transform(SOFT)method, theChebychevpropagationmethodandtheMulti
CongurationTimeDependentHartree(MCTDH)method. IntheSOFTmethod
theevolutiontimeisdividedintoanumberofslices. Foreachslice,thealgoritm
transforms back and forth between the momentum space and position space.
The potentialoperatoris treatedin the position space whilethe kinetic
oper-ator is treatedin themomentum space. If thenumber of slices is suciently
high, convergencecanbereached.
IntheChebychevmethod,theevolutionoperatorisexpandedin a
polyno-mialseriesintheoperator
exp[−i∆tH]
. Thebestapproximationisachievedby an expansionbased on complexChebychev polynomials [27]. The Chebychevmethod was used succesfully in paper I when the propogation wasperformed
on the ion-pair surface alone, but when couplings to the other states where
includedweencounterednumericalproblems and themethod wasabandoned.
Insteadweestimatedthelossofuxto(andregainingfrom)theothersurfaces
by calculatinga diabatic survivalprobability alongthe classicaltrajectory on
theion-pairpotentialusingtheLandau-Zenermodel[9]. WhentheChebychev
method was used in paper I, thecross section was obtainedby analyzing the
uxthrougaplaneintheasymptoticregion
(z = z
stop
)
verticaltothereaction coordinatez
.In paper II,we are using the MCTDH method instead of the Chebychew
method topropagatewavepacketsonthecoupledpotentialsurfaces,thistime
Figure 2.9: Wavepacketevolutiononthe
1
1
Π
resonantstateof
HF
+
IntheCranck-Nicholsonmethod[28]theevolutionoperatorisapproximated
withtheCayleyform
e
−
i∆tH
≈
1 − i
1
2
∆tH
1 + i
1
2
∆tH
(2.51)Thisapproximationisbothstableandunitary,incontrarytoexplicitmethods,
whereonlytermsuptorstorderisincludedin theexpansionoftheevolution
operator in a Taylorseries. This method was used in paperIII to propagate
wavepacketsontheresonantpotentialcurves. Figure2.9showsthetime
evolu-tionof theinitial wavepacket onthelowest
1
Π
resonantstateof
HF
using the Crank-Nicholsonmethod.When the wave packets have been propagated out in the asymptotic
re-giontheyareprojectedontoenergynormalizedwavefunctionsfortheseparated
atoms. Thisgivesus thetransitionmatrixelement. Fortheresonantstate
i
,it isgivenbyT
i
(E) = lim
t→∞
hΦ
E
(R)|Ψ
i
(R, t)i
R
,
(2.52)
andthecrosssectionforstate
i
isgivenby[29,22]σ
i
(E) =
2π
E
g|T
i
(E)|
2
,
(2.53)
σ
tot
=
X
i
σ
i
.
(2.54)ForthestatesofHFthatdissociateintoneutralfragmentsanddonothave
abarrierinthepotentialcurve,weuseplanewavesasenergynormalized
wave-functions:
Φ
E
(R) =
r
µ
2πk
e
ikR
.
(2.55)Here
µ
is thereducedmassandk
isthewavenumberrelatedto theenergyof thedissociatingfragments.Ifapotentialhasabarrierwefoundthatpartofthewavepacketgottrapped
beforethisbarrierand didnotreachtheasymptoticregion. Ifwethenproject
ontoplanewavesunrealinteferensepatternsshowupinthecrosssection. Also,
thecrosssectionwillbetimedependent. Instead,energynormalized
eigenfunc-tionsareusedintheprojectionoftheasymptoticwavepackets. Thesefunctions
werealsousedfortheion-pairstate,thathasalongrangeCoulombpotential.
The bound vibrational wave functions and the free energy eigenfunctions
˜
Φ
v
(R)
are calculated with the same method that was used to calculate the groundvibrationalwavefunctionforHF
+
. Thismethodalsogiveustheenergy
eigenvalues
E(v)
˜
. The energy normalized wave functions for these states are givenby:Φ
E
(R) =
d ˜
E(v)
dv
!
−
1/2
˜
Φ
v
(R).
(2.56) 2.3.2 The MCTDH methodThe MulticongurationTimeDependentHartree (MCTDH) programpackage
[30] is developed to study wavepacket dynamics in many dimensions. It has
succesfullybeenusedtostudyupto24degreesoffreedom[31]. InpaperIIthis
methodisusedtostudytheion-pairformationprocessinelectronrecombination
with
H
+
3
includingtwodimensions. ThebasicideaofMCTDHisthatthenuclear wavepacketcanbewrittenasasumofseperableterms. Intwodimensionthisbecomes
Ψ(t, r, z) =
n
r
X
i=1
n
z
X
j=1
A
ij
(t)ϕ
(r)
i
(r, t)ϕ
(z)
j
(z, t)
(2.57)where
A
ij
denotetheMCTDHexpansioncoecientsandϕ
(q)
i
arethe single-particlefunctions. IftheHamiltonianofthesystemcanbewrittenasasumofonedimensionaloperators,onlyoneterminequation(2.57)hastobeincluded
Ψ(t, r, z) = A(t)ϕ
(r)
(r, t)ϕ
(z)
(z, t)
(2.58)and the Time Dependent Hartree (TDH) method is obtained. In the TDH
method, the correlationbetween thetwonucleardegreesof freedomis not
and
n
r
, n
z
= 10
single particle functions are taken into account in equation (2.57).Eachsingleparticlefunction,
φ
(q)
i
isrepresentedintermsofaprimitivebasis:φ
(q)
i
(q, t) =
N
q
X
k=i
c
(q)
ik
χ
(q)
k
(q).
(2.59)Thetimeindependentbasisfunctionsareforcomputationaleciencychosenas
thebasisfunctionsofdiscretevariablerepresentation(DVR) [31].
Byinsertingtheansatz(2.57)intothetimedependentSchrödingerequation,
timedependentcoupleddierentialequationsfor thecoecients
A
ij
(t)
aswell as the single particle functionsφ
(q)
i
(q, t)
are obtained. The solution of these equations are not unique. Uniqueness is achieved by imposing constraints onthesingleparticlefunctionssuch askeepingthemorthonormalforalltime.
In order for the MCTDH method to be eective the Hamiltonian of the
system must be written as a sum of products of single coordinate operators.
This is typicallythe casefor the kineticenergy operator. Forexample, using
the same set of coordinates (Jacobi coordinates) as in paper II, the kinetic
energyoperatorhastheform
ˆ
T = −
2µr
1
∂
2
∂r
2
−
1
2µz
∂
2
∂z
2
.
(2.60)Alsothe potential energy surfaces,electronic couplings and autoionization
widthsmust bettedintothedesiredproductform
V (r, z) =
m
r
X
i=1
m
z
X
j=1
V
ij
f
i
(r)g
j
(z).
(2.61)Insection2.2.3,itwasdescribedhowthiswasperformedfortheion-pairsurface
relevantforpaperII.
Thewavepackets are propagated outin the asymptoticregion. Thecross
section for ion-pair formation and dissociative recombination is calculated by
integratingtheux absorbedbyacomplexabsorbingpotentialsituatedin the
Results and Discussion
3.1 Paper I
Inthis paper,the processof resonantion-pairformation(RIP) in electron
re-combination with
H
+
3
is studied. Weare performing structure calculationsinC
2v
symmetry since it had earlier beenfound that the resonantstates of the systemwerestronglyrepulsivewithrespecttobothJacobicoordinatesbutrel-ativelyatwhentheanglewasvaried.
Theion-pair state is identied by following the
(1a
1
)
2
(2a
1
)
2
conguration and atransformation from adiabaticto diabatic statesis performedusing thegeometry dependence of theCI coecients. The electronic couplingsbetween
theresonantstatesandbetweentheion-pairstateandthelowerRydbergstates
are obtainedusing atwo-by-twotransformationusingthe CIcoecients. The
electronic couplings betweenthe higherRydberg statesand the ion-pairstate
at small internuclear distances are estimated using a scaling of theelectronic
coupling between the ionization continuum and the ion-pair state. At large
internuclear distances the two-by-two transformation and the CI coecients
wereused to obtainthe electronic coupling betweentheion-pair stateand all
Rydbergstatesincluded.
The electroncapture induce wave packets on the tworesonant states and
autoionization isincludedby lettingthe potentialsbecomecomplexabovethe
ionpotentialsurface. Therealpartofthepotentialsurfaceistheenergyandthe
imaginarypart is negative with amagnitude given by half the autoionization
width. Thepositionsandwidths oftheresonantstateshadbeencalculatedin
anearlierstudy.
Thedissociationdynamicsontheion-pairsurfaceandonthetwocoupled
res-onantstatesaredescribedbysolvingthetimedependentSchrödingerequation
usingtheChebyshevpropagatorandthecrosssectionisobtainedbyanalyzing
theux throughaplaneat thereactioncoordinate
z = z
stop
. Toestimatethe luss of ux to the Rydberg states, the survivalprobability along theclassicalpathwascalculatedusingtheLandau-Zenermodel.
In an earlier one dimensional study of the reaction, a double peak
struc-turein thecrosssectionwasobtained. This structurewasnotobservedin the
measured cross section at CRYRING and wasdescribed asan inteference
yields across sectionwithamagnitudethat is approximatelysix timeshigher
thenthemesuredcrosssectionandwithapeakshiftedtowardssomewhatlower
interaction energiesrelative to the measured cross section. When the second
resonantstateisincluded, theamplitudeof thecrosssectionis increasedbya
smallamountwhentheChebyshevmethodisused. Whenthelossofuxtothe
Rydbergstatesis estimatedbytheLandau-Zener model, theamplitudeofthe
crosssectionbecomessmallerthenthemeasuredcrosssection. Whenincluding
ux recoveryat the second crossing,the cross sectionincreasesslightly but is
still smaller than the measured cross section and peaked at lower interaction
energies.
Theconclusionsarethat inorderto describethedynamicsofthisreaction,
at leasttwodimensionsneedtobeincluded. Theseconddimensionwillsmear
outthe interferenceeects. Toobtainacomplete pictureof thedynamics, all
calculatedstateshavetobeincludedin thewavepacketpropagation.
3.2 Paper II
Inthispaper,weusetheMultiCongurationTimeDependentHartree(MCTDH)
methodtostudythedynamicsoftheion-pairprocessinelectronrecombination
with
H
+
3
.Inorder tousethismethod thepotentialsurfaces,electronic couplingsand
autoionization widths need to be tted. The potential surfaces calculated in
paperI arettedtoatwodimensionalfunction ofMorse-characterandthe45
unknowncoecientsofthisfunctionwereoptimized. Theautoionizationwidths
aswellas theelectronic couplingsto theRydberg statescalculatedin paperI
are tted with products of gaussian functions. For theautoionization widths
andtheelectroniccouplingsbetweentheresonantstates,totally15coecients
are included. Since theelectronic couplings betweenthe ion-pair surface and
theRydbergstatesaremorespreadout,36coecientswereoptimized.
Thedissociateduxontheion-pairstateisanalyzedthroughacomplex
ab-sorbingpotential. Suchabsorbingpotentialsarealsousedfortheothersurfaces
to avoidreections towardsthe edge ofthe grid. Thecross sectionweobtain
usingtheMCTDHmethod,whenwepropagatethewavepacketontheion-pair
surfacealone,isshiftedtowardssmallerenergiesrelativetothecrosssectionwe
obtainwiththeChebychevmethod. ThecrosssectioncalculatedusingMCTDH
alsohasaslightlysmalleramplitude. TheMCTDHmethodisseveralordersof
magnitude faster and convergence is easierachieved. A draw backis that the
potentialsandwidths needto betted.
The cross section obtainedwhen awavepacket is propagated on the
ion-pairstatealoneusingtheMCTDHprogramisaboutafactorofsixlargerthan
the measured cross section. When the second resonantstate is included, the
crosssectiondropswith afactorof0.27. Whenthediabaticion-pairstatehas
crossedthe groundstateof theion, it will alsocross themanifoldof Rydberg
states below. We are hereincludingelectronic couplingsbetween theion-pair
stateandthefourlowestRydbergstates. Asexpected,uxwillbelostthrough
this interaction and cause the magnitude of the cross section to drop. When
the electronic couplingsto the Rydberg statesare included, themagnitude of