Reaction dynamics on highly
excited states
c
Johanna Brinne Roos, Stockholm 2009
ISBN 978-91-7155-869-5
Printed in Sweden by Universitetsservice US-AB, Stockholm 2009
Distributor: Department of Physics, Stockholm University
Abstract
In this thesis I have performed theoreti al studies on the rea tion
dy-nami s in few-atom mole ules. In parti ular, I have looked at rea tion
pro esses in whi h highly ex ited resonant states are involved. When
highlyex itedstatesareformed,the dynami sbe omes ompli atedand
approximationsnormallyusedin hemi alrea tionstudiesarenolonger
appli able.
To al ulate the potential energy urve for some of these states as a
fun tionofinternu lear distan e,a ombinationofstru ture al ulations
and s attering al ulations have to be performed, and the rea tion
dy-nami son the potentials hasbeen studied using both time-independent
andtime-dependent methods.
Thepro esses thathave been studied and whi h aredis ussed inthis
thesisareion-pairformationinele tronre ombinationwithH
+
3
,disso ia-tivere ombinationandion-pairformationofHF
+
,mutualneutralization inH+
+F−
ollisionsanddisso iativere ombinationofBeH
+
.Isotope
List of Papers
Thisthesisisbasedonthe followingpapers, whi h arereferredto inthe
text bytheir Romannumerals.
I Ion-pair formation in ele tron re ombination with
H
+
3
Å.Larson,J.RoosandA. E. Orel
Phil. Trans. R.So . A,364,2999 (2006)
II Ele tron ollisions with
H
+
3
: ion-pair formationJ.B.Roos,Å.Larson and A.E. Orel
Phys.Rev. A,76, 042703(2007)
III Disso iative re ombination of
HF
+
J.B.Roos,Å.Larson and A.E. Orel
Phys.Rev. A,78, 022508(2008)
IV Resonant ion-pair formation in ele tron
re ombina-tion of
HF
+
J.B.Roos,A.E. OrelandÅ. Larson
(Manus ript, submitted to Phys.Rev. A(2009))
V Mutual neutralization of
H
+
+ F
−
J. B. Roos, J. Zs. Mezei, K. Shilyaeva, N. Elander and Å.
Larson
(Manus ript, to be submittedto Phys.Rev. A(2009))
VI Disso iative re ombination of
BeH
+
J.B.Roos,M. Larson, Å.Larsonand A.E. Orel
(Manus ript, submitted to Phys.Rev. A(2009))
VII Ion-pair formation in ele tron re ombination with
mole ular ions
Å.Larson,J.B. Roos, M. Stenrupand A.E. Orel
J.Phys.: Conf. Ser.,88, 012065(2007)
Publications not included in the thesis
VIII A dire t, lo al model of disso iative re ombination
of
HF
with
H
+
3
Å.Larson,J.B. Roosand A.E. Orel
A epted byJ.Phys.: Conf. Ser.(2009)
X Resonan es in disso iative re ombination: trends
and patterns
A.E.Orel, Å.Larson,J.B.Roos,V.NgassamandJ.Royal
A epted byJ.Phys.: Conf. Ser.(2009)
Contributions by the author
I Iperformedthe stru ture al ulations, arriedoutthe
diaba-tizationofthe al ulatedadiabati potential energysurfa es
anddeterminedtheele troni ouplings.Ialso al ulatedthe
lassi al rea tion path on theion-pair potential energy
sur-fa e whi h is then used for the semi lassi al al ulation of
theion-pairrea tion rossse tion.
II The multi onguration time dependent Hartree method
used to propagate the wave-pa kets in this paper requires
that all potential energy surfa es and ouplings are in
produ t form. I have developed a fun tion that has the
required form and works well for all of the diabati
potentials al ulated earlier and I then tted these
potentialsto thatfun tion.Ihave alsoperformedthetting
of all ele troni ouplings and autoionization widths to
appropriateforms.
III Iperformed all al ulations and wrotethepaper.
IV Iperformed all al ulations and wrote thepaper. In
parti -ular,Idevelopedadiabatizationmethodinwhi hboth
qua-sidiabati potential energy surfa es and ele troni oupling
are al ulated simultaneously.
V Imodiedthe al ulations usedto studymutual
neutraliza-tion pro esses, performed al ulations using the potentials
and ouplings we reported in paper IV and analyzed the
stru turesinthe rossse tion.
VI I performed the diabatization of the al ulated adiabati
potential energy surfa es and al ulated the ele troni
ou-plings.I also performedthe dynami al al ulations and
al- ulated the rossse tion.
Contents
1
Introduction
. . . .
13
1.1 Dissociative recombination
. . . .
131.2 Resonant ion-pair formation
. . . .
171.3 Mutual neutralization
. . . .
192 Theory
. . . .
23
2.1 Potential energy curves and couplings
. . . .
232.1.1
Approximations and definitions
. . . .
232.1.2
Structure calculations
. . . .
282.1.3
Scattering calculations
. . . .
312.1.4
Diabatization
. . . .
402.1.5
Extrapolation
. . . .
452.2 Reaction dynamics
. . . .
462.2.1
The driven Schrödinger equation
. . . .
462.2.2
Wavepacket dynamics
. . . .
542.2.3
The log-derivative method
. . . .
612.2.4
Identifying resonances
. . . .
623 Results and Discussion
. . . .
65
3.1 Paper I
. . . .
653.2 Paper II
. . . .
673.3 Paper III
. . . .
693.4 Paper IV
. . . .
713.5 Paper V
. . . .
753.6 Paper VI
. . . .
783.7 Paper VII
. . . .
1. Introduction
In this thesis I have studied rea tions that are of importan e for the
physi s of plasmas where mole ular ions are present. There isa la k of
understanding of how these ions are formed, ex ited and destroyed in
elds su h as astrophysi s and plasma-based te hnologies, and this has
been a strongmotivationfor this resear h.
The abundan e of H
+
3
indiuse mole ular louds,for example,trou-bled resear hers for de ades until Kokoouline et al. in 2003 in luded a
previously overlooked mole ular ee t in their theoreti al treatment of
the disso iative re ombination of H
+
3
[1℄ and their new results yieldedgoodagreement between theory andexperiments.
In et hing plasmas, undesirable rest produ ts are sometimes reated
that an be harmful to the environment [2℄ and to prevent this from
happeningitisimportanttoknowmoreaboutthepro esseswithinthese
plasmas.
The disso iative re ombination (DR), resonant ion-pair formation
(RIP) and mutual neutralization (MN) rea tions are similar in many
ways and in the remaining part of this hapter I will give a short
introdu tion to thesepro esses.
1.1
Dissociative recombination
Disso iativere ombination(DR)isthepro esswhereamole ular ation
re ombines with an ele tron to form a highly ex ited neutral mole ule
thatdisso iatesinto neutral fragments.
One might ask oneselfhow it an be possible for an ele tron, whi h,
relativetothesmallestmole ularion,isthesizeofanant ompared toa
elephant,tohavesu hanimpa tonamole ule?Su hapossibilityexists,
however,wheneverthe times ale oftherea tion orrespondstothatof a
mole ular vibration of the ion. The ele tron an then atta h to theion
to formahighly ex itedneutral state thatisstrongly repulsive,a
prop-ertythatimmediatelyfor esthemole uleto disso iate.Thedisso iation
pro ess is fast ompared to ompeting pro esses su h as relaxation by
photonemissionand thereforedominates.
Theneutralintermediatestateformedinthisrea tionis alleda
reso-nantstate.Itisnotstableinthetraditionalsense,sin e therealsoexists
theneutral mole ule reemits the ele tron ba kto the ionization
ontin-uumandfallsba kintoavibrationalstateoftheion.Thispro esso urs
on the same times ale asdisso iation and also needs to be onsidered.
Theprobability forele tron apture into andautoionization froma
par-ti ular state is related to the autoionization width, whi h is inversely
proportional to the lifetime of this state. When the resonant state has
rossedtheiongroundstate,autoionizationisnolonger possibleandthe
resonant state be omesenergeti allystable towardsautoionization.
A possible disso iation path is illustrated inFig. (1.1). Here, theion
islabeledAB
+
,theresonantstateAB
∗∗
andtheresultingDRfragments
atlargeinternu lear distan esarelabeledwithA+B.Theprobabilityof
autoionization from the resonant state is indi ated by adding a width
to theresonant state potential above the ionstate potential. In theory,
however,theautoionizationwidthisrepresentedasa omplexpartofthe
state,whi hiszerowhenthestateisele troni allystable.Thehorizontal
line indi ates the ground vibrational state of the ion where the pro ess
might start, andthe arrowsindi ate thedire tion oftherea tion.
Fig. (1.1) is a very simplied pi ture of a DR event. In general the
initial ondition of the ion is more orre tly des ribed by some kind
of distribution among of vibrational and rotational states. There might
an innite number of Rydberg states whi h onverge to this state. A
Rydbergstate hasthe same ongurationasthe ionplusan ele tronin
an outerorbital. Resonant states areRydbergstates whi h onvergeto
ele troni ally ex ited states of the ion, and sothere are also an innite
numberofthese.However,belowa ertainenergy,thenumberislimited
althoughmight still be large.
In1950,SirDavidBateswastherstpersontosuggestthatthede ay
inthelevelofionization intheionosphere ouldbeexplainedbyDR[3℄.
In order to do so, he used a diabati representation. In a diabati
rep-resentation states of the same symmetry are allowed to ross, and are
oupled to ea h other by ele troni ouplings. In this diabati pi ture,
and in ea h symmetry, we an have a large number of resonant states
rossing an innite number of Rydberg states. All of these states are
oupledto ea h otherbyele troni ouplings.
TherearetwomodesofDR,adire tandanindire tone.Intheformer,
rst proposed by Bates [3℄, the ele tron is resonantly aptured by the
ionand losesits energy to ele troni ex itation of the resulting neutral
mole ule. This is the pro ess illustrated by the potential energy urves
showninFig.(1.1).Inthe latterpro ess,theindire tmodeproposedby
Bardsley[4℄, theele tron looseitsenergy to rovibroni ex itationofthe
resulting neutral mole ule.In otherwords, the ele tronis aptured into
aRydbergstate andifthisstate then rossesarepulsive state,
predisso- iationispossible.Predisso iationisenhan ed iftheele troni oupling
between the rossing states is large. This is the pro ess illustrated in
Fig.(1.2).Here, the Rydbergstate is labeledbyAB
∗
.
Thedire tandindire tmodeoftheDRpro ess anberepresentedby
AB
+
+
e−
→
AB∗∗
→
A+
B (1.1) AB+
+
e−
→
AB∗
→
AB∗∗
→
A+
B (1.2)respe tively. These two pro esses o ur simultaneously and
ompeti-tion between them has to be onsidered for a orre t des ription [5℄.
InFig.(1.1)andFig.(1.2) onlyonedisso iationpathisdisplayed.
How-ever, as mentioned earlier, there exist an innite number of Rydberg
states below the ion and there an also be a large number of resonant
statesthat rosstheionpotential.Allofthestatesthatareenergeti ally
open for disso iationat a given energy and asymptoti allyformneutral
fragments areavailable DR hannels and ux from resonant states an
be redistributed among all of the states that ele troni ally oupled to
these.Byenergeti allyopenfordisso iation, Imeanthattheintera tion
energy between the ion and the ele tron, measured relative theground
vibrational state of the ionwhere the rea tion starts, is larger than the
There an alsoexist statesthat asymptoti ally formioni fragments,
andthispro essisex ludedfromDR.Whendisso iationintothis
han-nel happens, the pro ess is referred to as resonant ion-pair formation
(RIP).Thisrea tion willbe dis ussedinthe following se tion.
For simpli ity, Ihave only onsidered a diatomi mole ule ABwhere
thepotential energy urvesonly depend on one internal oordinate,the
internu lear distan e.Inthe aseofpolyatomi mole ules,thedynami s
takespla eonmultidimensionalsurfa esandthe fragmentsAandB an
be mole ular or atomi .
Weinvestigatearea tionby al ulatinga rossse tion.A rossse tion
isa measure of the probability for a rea tion to o ur and isan energy
dependent quantity. For an ele tron re ombination pro ess su h asDR,
inthesimplest ase theenergy isgiven bythe ollision energy between
theinitial ele tronandthemole ular ion.Fluxthatinitiallyis aptured
into the resonant states and that is not lost due to autoionization will
be redistributed among all states below the ion. The ross se tion for
DR is the sum of partial ross se tions of all neutral hannels. Sin e
theabsolutemajorityof statesbelowtheionarestatesdisso iatinginto
neutralfragments,theex eptionbeingtheion-pairstate,itisreasonable
toassumethata rossse tionforDR al ulatedbyex ludingele troni
nalstate distributions are to be determined, it is of great importan e
to in lude theele troni ouplingsbetween the states.
When the ele tron re ombines with the mole ular ion there is a
Coulomb intera tion between theion andtheele tron. It wasshown by
E. P. Wigner in 1948 that for a Coulomb attra tion between olliding
parti les, at low ollision energies the ross se tionfor the rea tion will
be inversely proportionalto the energy [6℄.
From this short introdu tion of DR itis not hard to understand that
thetheoreti altreatment ofthispro essisfarfromtrivial.Inthetheory
se tionofthisthesis,theapproximationsandmethodsusedfor
perform-ingthese al ulationswill be dis ussed.
1.2
Resonant ion-pair formation
Resonant ion-pair formation (RIP) is a similar pro ess to DR ex ept
that oppositely harged pairs of fragments are formed. The dire t and
indire tmode of RIP an be representedby
AB
+
+
e−
→
AB∗∗
→
A+
+
B−
(1.3) AB+
+
e−
→
AB∗
→
AB∗∗
→
A+
+
B−
(1.4) respe tively.The ele tron anity of an atom or mole ules is the energy released
whenthe extraele tron isdeta hed froma singly hargednegative ion.
Assumethatthemole ularionAB
+
disso iatesintofragmentsA
+
+
B.If
theele tron anityofthe Bfragment intheA
+
+
B
−
hannelissmaller
thanthedisso iationenergyoftheinitialion,theasymptoti limitofthe
ion-pairstate liesabove thegroundvibrationalstate oftheionand this
hannel is then energeti ally losed for disso iation at zero intera tion
energy.ThisisillustratedinFig.(1.3). In ontrastto theneutralstates
belowtheion, the ion-pairstateis notat intheasymptoti region due
to the Coulomb attra tion between the ion pair, and this is shown in
Fig.(1.3).
Dierentexperimentalte hniqueshavebeen developed overtheyears
tostudyrea tionssu hasDRandRIP.Oneofthemostsu essful
meth-ods inuseis theionstoragering[7℄.Here, themole ular ionsarestored
underhigh-va uum onditionsathighkineti energiessothatI-Ra tive
ions have time to relax down to the ground vibrational state through
photon emission. In this way, a well dened initial ion-state is formed.
Theionsarethen ollidedina ontrolledmannedoverasmallintera tion
region with old ele trons ina o-parallel merged beams onguration.
The bending magnets in the storage ring make it easy to separate the
Figure 1.3: S hemati potentialenergy urvesfortheRIP pro ess. D
0
(AB+
) isthedisso iationenergyfortheionandEA(B)istheele tronanityforB
alsoawelldenednalstatefortheRIPpro ess.StudyingRIPpro esses
theoreti ally and omparing results with experimental measurements is
ne essary if we are to rea h a deeper understanding of the underlying
quantum ee ts that ontribute to the stru ture whi h an be seen in
therea tion rossse tion.
Stru tures in the ross se tion ould arise from, for example,
interferen e between dierent pathways and tunneling whi h are both
purelyquantum me hani al ee ts. These phenomena areillustrated in
Fig. (1.4). It is the oupling between a Rydberg state and a resonant
state that makes it possible at the rossing point for ux whi h is
initially in the resonant state to ouple out onto the Rydberg state. If
the same urves ross again, the pro ess an o ur again and due to
the fa t that there are now several ompeting pathways to rea h the
same point, interferen e between the various ux pa kets will o ur.
Without su h rossings, tunneling through the barrier between states
anstill o ur. Stru turesinthe rossse tionthataredue to tunneling
pro esses are alled shape resonan es and stru tures with their origin
in interferen e between dierent pathways are alled Stü kelberg [8℄
se tion.
By omparing our al ulated RIP ross se tions with those
experi-mentally measured in ion storage rings, su h as CRYRING in
Sto k-holm [9,10℄, and TSR inHeidelberg [11, 12℄ we have rea hed a deeper
understandingintheimportan eofhaving welldes ribedpotentialsand
ouplingsinour al ulations.
1.3
Mutual neutralization
Mutualneutralization (MN)o urs onthesame set ofpotentialsasthe
ele tronre ombinationpro essesDRandRIP.Butinthis ase,the
rea -tionstartsoutintheion-pairlimitandgoestowardssmallerinternu lear
distan es. An ele tron is transferred between the negative and positive
ions,and after ree tion towards theinner walls of the potentials,
neu-tral fragments are formed. In a quasidiabati pi ture, where the
ele -troni ouplingsarelo alizedatthe urve rossings,anele trontransfer
eventismostlikelytoo urattheinterse tionbetweenthe ion-pairand
neutral state potentials.
TheMN pro essis illustratedinFig.(1.5)and anberepresentedby
A
+
+
B
−
The neutral fragments formed inthe rea tion might be in formed in
their ground or ex ited states. Note that if the ion-pair state does not
rossthe groundstate, there is no a essto this state, and only one of
theneutral fragments an be formed in its ground state,see Fig. (1.5).
From thepotentials in this gure we an see that in this ase, the rst
hannel, the ground state A(1)+B(1), is not available sin e there is no
rossingand hen e no a essto theex ited states. The se ond hannel,
A(2)+B(1), isopen at zero ollision energy sin e this limit is belowthe
ion-pair limit A
+
+
B
−
and so the ion-pair state must ross this state
twi esothereisanenhan edpossibilityofinterferen ebetweendierent
pathways for this hannel.
To rea hthethirdand forth hannel, A(3)+B(1)and A(4)+B(1),the
ollision energy must be higher than their orresponding threshold
en-ergies E
th
(1)
and Eth
(2)
. This means that the ross se tions for these hannels arezero for energiesbelowEth
(1)
and Eth
(2)
respe tively.In summary, the shape of the ross se tion for a MN rea tion very
mu hdependsonthepositionoftheion-pairstaterelativetotheneutral
states.However,quantumee tssu hastunnelingandinterferen es will
also ontributewith moreor lesspronoun ed stru tures.
The rea tion of H
+
+ H
−
has served as a ben hmark for MN studies
mid-fties by Bates and Lewis [13℄and sin e then more rigorous
treat-ments of the rea tion have been published [14,15, 16℄. This parti ular
rea tion ano urat zero ollisionenergy sin einthis asetheion-pair
limitlies above someof theneutral hannels.
InpaperV,we studytheMN ofH
+
+
F
−
.Our al ulations showthat
an ex ess energy is needed for the rea tion to o ur, i.e. at non-zero
ollisionenergies. However, otherstru ture al ulations ofthe potential
energy urvesof HFshowan interse tionbetween theground state and
theion-pair state,and ifthis is the ase, no ex essenergy isneededfor
thisrea tion to o ureither.
Finally,notethattheautoionizationpro esseshavenotbeenin luded
in our theoreti al treatment of MN. In the example urves shown in
Fig.(1.5) theresonant state liesabove theionAB
+
urve and an then
be rea hed even at zero ollision energy if the ion-pair limit is above
the ground vibrational state of the ion. Autoionization may therefore
ontributeandour theoreti almodelhastobedeveloped toin lude this
possibility.
DESIREE is a double ele trostati storage ring whi h is now
un-der onstru tion in Sto kholm [17℄. It is designed for experiments with
merged beams of positive and negative ions and one of the aimsof the
apparatus is to allow mutual neutralization rea tions to be studied at
low ollision energies. Using an imaging te hnique [18℄ it will also be
possible to determine the nalstate distributions of the produ ts from
MNrea tions.Comparingour al ulatedMN rossse tionswiththe
or-responding experimental rossse tion measured with DESIREE would
provide us withmoreinformation onthevalidity ofthe approximations
2. Theory
In this hapter, I will dis uss the theoreti al ba kground to the papers
in luded in the thesis. For simpli ity and larity in the derivations of
the equations, I will des ribe the ase of a diatomi mole ule with an
internu lear distan e
R
whenthe equations inthis hapter are derived.Coordinates of all ele trons are represented by
r
.The generalization topolyatomi mole ules, where
R
be omesave tor,isstraightforward buttheequations aremore ompli ated. Atomi units are usedthroughout
this thesis unless otherwise stated, i.e.,
~
= k
e
= m
e
= e = 1
,where~
is the redu ed Plan k onstant,
k
e
is the Coulomb onstant,m
e
is theele tronmassand
e
istheelementary harge.2.1
Potential energy curves and couplings
Inthisthesis, the out ome ofa hemi alrea tionis investigatedbyrst
al ulatingpotential energy urves and ouplings for themole ular
sys-temand thenstudyingthe nu lear dynami s onthese urves. Thebasis
fortheideathatnu leardynami stakespla eonpotentialenergy urves
isthefamousBorn-Oppenheimer approximation[19℄,inwhi hthe
time-independentS hrödingerequationofamole uleisseparatedintoan
ele -troni andanu lear part.Apotentialenergy urveisthen al ulated by
solvingtheele troni S hrödingerequationatdierentxedinternu lear
distan es.In order to go beyond theBorn-Oppenheimer approximation
dierent representations of the ele troni states of a mole ule an be
used.Themost ommon istheadiabati representation, wherethe
ele -troni Hamiltonianisdiagonalized.For somepurposeshowever,abetter
option an be to usea non-diagonal diabati representation of the
ele -troni states. In a diabati representation, the diagonal elements of the
ele troni Hamiltonian are the diabati potentials and the o-diagonal
elementstheele troni ouplingsbetween thesestates.
2.1.1
Approximations and definitions
Thenon-relativisti Hamiltonianof adiatomi mole ule hastheform
H = −
2µ
1
∇
2
R
+ H
el,
H
el= −
1
2
n
elX
i=1
∇
2
r
i
+
Z
A
Z
B
R
+
n
elX
i=1
n
elX
j>i
1
r
ij
−
n
elX
i=1
Z
A
r
Ai
+
Z
B
r
Bi
.
(2.2)The rst term in Eq. (2.1) is the nu lear kineti energy operator
T
N
(R, θ, ϕ)
, and this an be separated further into a rotational and avibrational term whi h des ribe therotation of thenu lear axisaround
the enter-of-mass of the mole ule and thevibrations along thenu lear
axis, respe tively. The internu lear separation ve tor
(R, θ, ϕ)
denotesthespatial orientation ofthemole ule.The rstterminEq. (2.2)isthe
ele tron kineti energy operator
T
e
(r)
. The remaining terms are theele trostati potential energy,
V (R, r)
, onsistingof thenu lear-nu learand ele tron-ele tron repulsions and the ele tron-nu lear attra tion.
µ
istheredu ed massofthetwo nu lei,
A
andB
,andisdened asµ =
M
A
M
B
M
A
+ M
B
,
(2.3)and
Z
A
andZ
B
arethe nu lear hargestates.To nd solutions
ψ(R, θ, ϕ; r)
to the time-independent S hrödingerequation
Hψ(R, θ, ϕ; r) = Eψ(R, θ, ϕ; r),
(2.4)thetotal wavefun tion an beapproximated bythe produ tof an
ele -troni anda nu lear wave fun tion
ψ(R, θ, ϕ; r) = χ(R, θ, ϕ)φ(R, r).
(2.5)Thisistheso alledBorn-Oppenheimerprodu tand isinsertedintothe
time-independentS hrödingerequation, Eq.(2.4).Usingtheassumption
thatthe ele trons move mu h faster than the mu h heavier nu lei, this
equation an be separated into a nu lear and an ele troni part. The
ele troni S hrödinger equation for adiatomi mole ule hastheform
H
el
φ
a
i
(R, r) = E
i
a
φ
a
i
(R, r),
(2.6)where
R
no longer isa variablebut a parameter.By solving thisdiago-nalequationatdierentnu lear positions,anadiabati potentialenergy
urve
E
a
i
(R)
is formed.The potential energy urve is alled adiabatisin eitisassumedthattheele tronsrespondimmediatelytothemotion
ofthe nu lei.The nu lear time-independent S hrödinger equation
[T
N
+ E
i
a
]χ
a
i
(R, θ, ϕ) = E
i
χ
a
i
(R, θ, ϕ),
(2.7)des ribes the nu lear motion on the state
i
. It an be the vibrationalorrotational motionofbound ele troni statesand alsodisso iation
tenagood approximationfor ele troni ground statessin e thesearein
generalwellseparatedinenergyfrom theex ited states.For theex ited
states,ontheotherhand,thepotentialenergy urvesarenotalwayswell
separatedfromea hotherandthestatesintera tmoreeasily.Inthese
re-gions,nu leardynami swillnotfollowtheele troni statesadiabati ally
andtheBorn-Oppenheimer approximationbreaksdown.
The intera tion of the nu lear and ele troni motion is responsible
both for the ouplings between the dierent neutral ele troni states
andalsofor autoionization. When des ribingpro essessu h as
disso ia-tive re ombination, ion-pairformation,and mutualneutralization,these
ouplings are ru ial and soit is absolutely ne essaryto go beyond the
Born-Oppenheimer approximationifthese rea tions areto be studied.
Inonesu hmodelwhi hgoesbeyondtheBorn-Oppenheimer
approxi-mation,theadiabati stateswillget oupledbythekineti energy
opera-tor
T
N
(R, θ, ϕ)
.Asmentionedearlier,thenu learkineti energyoperatoranbe dividedinto a vibrationaland a rotationalpart
T
N
(R, θ, ϕ) = T
vib
(R) + T
rot
(θ, ϕ),
(2.8) whereT
vib
(R) = −
1
2µR
2
∂
∂R
R
2
∂
∂R
(2.9) andT
rot
(θ, ϕ) = −
1
2µR
2
1
sin θ
∂
∂θ
sin θ
∂
∂θ
+
1
sin
2
θ
∂
2
∂ϕ
2
.
(2.10)Assume that Eq. (2.6) has been solved for all
R
and that a full setof adiabati ele troni states have been obtained. The solutions an be
takenasorthogonalandnormalized andassu htheyprovidea omplete
basisforfun tionsdenedovertheele troni spa e.Thetotalwave
fun -tion an thenbeexpanded as
ψ(R, θ, ϕ; r) =
∞
X
i=1
χ
a
i
(R, θ, ϕ)φ
a
i
(R, r),
(2.11)where the nu lear wave fun tions
χ
a
i
(R, θ, ϕ)
have been in ludedin order to span the whole onguration spa e. If this expansion is
inserted in Eq. (2.4), and that expression is multiplied from the left
with
φ
a
j
(R, r)
and integrated over the ele troni oordinates a oupledtime-independent S hrödinger equation for the adiabati states an be
obtained
[T
N
− E
i
a
]χ
a
i
−
X
j
1
µ
F
a
ij
· ∇
R
+
1
2µ
G
a
ij
χ
a
j
= Eχ
a
i
,
(2.12)Internuclear distance
Potential energy
R
x
φ
a
1
,
φ
d
1
φ
a
2
,
φ
d
2
φ
a
1
,
φ
d
2
φ
a
2
,
φ
d
1
Figure 2.1: S hemati potentialenergy urves lose toa rossing. Solid lines are adiabati potential energy urvesand dashed lines are diabati potential energy urves.
φ
a
and
φ
d
representtheadiabati and diabati ele troni wave fun tionsrespe tivelyand
R
x
istheavoided rossingpoint.where
F
ij
a
= hφ
a
i
|∇
R
|φ
a
j
i
(2.13)G
a
ij
= hφ
a
i
|∇
2
R
|φ
a
j
i .
(2.14)Theo-diagonalelements
F
a
ij
andG
a
ij
a tas ouplingsbetweendierentele troni statesandarereferredtoasrstderivativeandse ond
deriva-tive non-adiabati ouplingsrespe tively.
F
a
ij
ispurelyo-diagonal whileG
a
ij
alsohasdiagonalterms,whi haresometimesreferredtoasadiabatiorre tions. Theo-diagonal ontributions from
G
a
ij
arenormally mu hsmallerthanthoseof
F
a
ij
andarethereforeoftennegle ted. Itwasshownin 1929 by Neumann and Wigner that the adiabati potential energy
urves,for adiatomi mole ule, orrespondingtoele troni statesofthe
samesymmetry annot ross[20℄.Thisisreferredto asthenon- rossing
rule.Instead,two adiabati potential energy urvesofthesame
symme-tryrepelea hotherwhenthey ome lose.Thepointof losestapproa h
is alledan avoided rossing point and inFig.(2.1) isindi ated by
R
x
.Thestrongest ontributiontothenon-adiabati ouplingintheregion
ofanavoided rossing omesfromtheradialpartof
F
a
se ond derivative oupling elements a ton
χ
a
i
. This type of dierentialismore di ult to solve than one ontaining ouplingswhi h have only
potentialform. Furthermore,thenumeri alevaluationofthese ouplings
analso benontrivial.
These issues an be ir umvented byusing diabati states dened in
su h a way that the rst derivative ouplings disappear. The rst
re-ported use of su h states, that later would be referred to as diabati
states, or a tually quasidiabati states, was by Zener in 1932 [21℄. The
adiabati ele troni eigenfun tions have a typi al hange of hara ter
lose to an avoided rossing. Zener assumed that the basis fun tions
used were approximative eigenfun tions of the ele troni Hamiltonian
without this hara teristi , and this is also indi ated in Fig (2.1). As
the name diabati suggests, these states do not adjust adiabati ally to
hanges intheinternu lear distan e.
ThestatesusedbyZenerarereallyquasidiabati ,sin etherst
deriva-tiveradial ouplingsdonotvanish ompletelywiththisapproa h.Ashas
been pointedout byC.A.Meadand D.G.Truhlar[22℄,every omplete
setof ele troni statesfullling
F
ij
d
(R) = hφ
d
i
|
∂
∂R
|φ
d
j
i = 0
(2.15)mustbeindependentofR.However,thebasissetusedina al ulationis
both nite and in omplete and thena diabatization within thenumber
ofstates onsidered anbeperformed. Thefollowingsteps arerequired.
First,thediabati statesare expandedintheadiabati statesas
φ
d
i
(R, r) =
M
X
j=1
φ
a
j
(R, r)T
Tij
(R) ,
(2.16)where
T
ij
is an element in the orthogonal adiabati to diabatitrans-formation matrix (ADTM) and
M
isthe number of adiabati statestobetransformed.Theexpansionistheninsertedintothestri tlydiabati
ondition,Eq. (2.15),andaftersome manipulations thefollowing
dier-ential equation for the ADTMmatrix is obtained
d
dR
T
+ F
a
T
= 0 .
(2.17)
This dierential equation an be solved with the following boundary
ondition
lim
R→∞
T(R) = 1,
(2.18)
whi hmakestheadiabati anddiabati potentialenergy urvesidenti al
furthermore, that the rst derivative ouplings between the adiabati
stateswith
i < M
andthosewithi≥M
arezero,it anbeshownthatthese ondderivative ouplingsbetweenthediabati statesarealsozero[23℄.
The oupledS hrödingerequationinthestri tlydiabati representation
be omes
T
N
χ
d
i
−
X
j
V
ij
d
χ
d
j
= Eχ
d
i
,
(2.19)where thediabati potential matrix elementsaregiven by
V
ij
d
= hφ
d
i
|H
el
|φ
d
j
i .
(2.20)In ontrast to theadiabati potential matrix,thediabati potential
ma-trix is not diagonal. The diagonal elements are the diabati potential
energy urves
E
d
i
(R) = V
ii
(R)
. The o-diagonal elements will ouplethe diabati ele troni states. These ouplings are alled the ele troni
oupling elements
c
ij
(R) = V
ij
(R), i 6= j
.Therelationship between theadiabati and the diabati potential matrix isgiven by
V
d
= T
TV
a
T
.
(2.21)In the ele tron re ombination pro esses that have been studied in the
work presented here, there exist an innite number of Rydberg states
onverging to the ion ore. Furthermore, there are a large number of
resonant states that are embedded in the ontinuum of the ion plus a
free ele tron. In general, these resonant states will ross both the
ion-ization ontinuum and the Rydberg states and this situation makes a
stri tdiabati representationimpossibleto use.Toresolvethis problem,
a kind of quasidiabati representation, in whi h theele troni oupling
elementsarelo alized totheregion of avoided rossings,hasbeen used.
The methods that have been employed to determine the quasidiabati
potentials andthe ouplings will be des ribed inalater se tion.
2.1.2
Structure calculations
Sin e the states involved in the pro esses under investigation onsist
of both ele troni ally stable states belowthe ion and ele troni ally
un-stable, autoionizing states, above the ion, the potential energy urves
annot be al ulated ompletely with the quantum hemistry methods
normally used for this purpose. S attering al ulations need to be
per-formedonthosestatesabovetheiontodeterminethetruepositionsand
theautoionizationwidthsoftheresonantstates.Belowtheion,however,
whereallstatesareele troni allystable, thepotentialenergy urves an
be al ulated using onventional stru ture al ulations. In the
s attering al ulations withthose obtained from stru ture al ulations,
thesamebasisset,typeoforbitals andquantum hemistry methodthat
has been used for the target wave fun tion, must also be used in the
stru ture al ulations on the neutral ele troni states. To obtain
a u-rateresults from the stru ture al ulations agoodbasisset and a large
expansion of the wave fun tion are needed. Unfortunately, due to the
s attering method used,thereis arestri tionon thesize ofthebasisset
andtheexpansionthat anbeusedandthismeant thatthe al ulations
on the potentials of the stable states were less a urate. However, the
benet ofthes attering results ompensates for this disadvantage sin e
it provides both the autoionization width and a mu h more a urate
resultfor theposition oftheresonant states.
Thequantum hemistry methods [24,25℄thathave been used to
al- ulate thepotentials of the stablestates inthis thesis will bedes ribed
hereand thes attering al ulationsare des ribedinthe nextse tion.
IntheHartree-Fo k(HF)method,alsoreferredtoastheself- onsistent
eldmethod,theele troni wavefun tion isexpressedasaSlater
deter-minant
φ
SD
=
1
√
N
el
!
ξ
1
(1)
ξ
2
(1)
. . .
ξ
N
el
(1)
ξ
1
(2)
ξ
2
(2)
. . .
ξ
N
el
(2)
. . . . . . . . . . . .ξ
1
(N
el
) ξ
2
(N
el
) . . . ξ
N
el
(N
el
)
,
(2.22)where
ξ
i
(j)
isthemole ularspinorbitali
forele tronj
.Thespatialpartsof the mole ular spin orbital onsist of mole ular orbitals (MOs) that
are onstru ted by linear ombinations of atomi orbitals. The optimal
wave fun tionisobtainedbyminimizing thefollowingexpressionforthe
energy
E
SD
= hφ
SD
|H
el
| φ
SD
i .
(2.23)Otherele troni stru ture al ulationsbeginwithaHartree-Fo k(HF)
al ulation.Fromthis,the mole ular orbitals(MOs)anda groundstate
solution an be obtained though ele tron orrelation ee ts are not
a - uratelyin luded.The ongurationintera tion(CI)methodisthe
sim-plestwaytoin ludethis orrelationandtothereforeimprovethe
Hartree-Fo ksolution.Thisisamethodwellsuited for al ulatingex itedstates
of small mole ules. The wave fun tion is then onstru ted as a linear
ombinationof Slater determinants
φ
CI
= φ
SD
+
X
i
c
i
φ
i
,
(2.24)where
φ
SD
is the initial Hartree-Fo k wave fun tion andφ
i
are Slaterthesolutionto the Hartree-Fo kequation.
For alineartrialwave fun tion,thevariational prin iplenowleadsto
solving the se ular equations for the CI oe ients. In this ase there
willbeasmanysolutionsasthereare ongurationsintheCIexpansion.
The solution with the lowest energy is the ground state and the other
solutions orrespondto ex ited states.
The eigenfun tions of the ele troni Hamiltonian are simultaneously
eigenfun tionsof the spinoperator and soanalternative approa h isto
use ongurationalstatefun tions(CSFs),whi harespinadaptedlinear
ombinationsofSlater determinantsdesignedtobeeigenfun tionsofthe
S
2
operator inthe CIexpansion.In a full CI (FCI) al ulation a omplete set of determinants is
gen-erated by distributing the ele trons among all of the orbitals and then
in ludedintheCIexpansion,Eq.(2.24).ThenumberofSlater
determi-nantsin reases very rapidly both withthenumberofele trons and also
withthenumberof orbitals. A FCIexpansion istherefore suitable only
for thesmallest ele troni systems.
For larger ele troni systemsthe FCIexpansion has to be trun ated.
When designing smaller onguration spa es, it is important to
dis-tinguish between stati and dynami orrelation. Stati orrelation is
treated by retaining the dominant ongurations of the FCI expansion
aswellasthosethatarenearlydegeneratewiththedominant
ongura-tions.These ongurationsarereferredtoasthereferen e ongurations
of theCI wave fun tion, and they span thereferen e spa e. Dynami al
orrelation is treated by adding ongurations whi h are generated by
ex itations outof the referen espa e to the wave fun tion.
Thereferen espa einamultireferen eCI(MRCI) al ulation should
ontain all those ongurations thatmaybe ome importantto des ribe
the potential energy urves needed to model a ertain rea tion. The
MRCIwave fun tion isgenerated byin luding all ongurations inthis
referen espa easwellasthose ongurationsgeneratedbyex itationsof
thereferen eele trons intothe virtualorbitals. Itis ommonto in lude
all singleand double ex itations out of thereferen e spa e, resulting in
aMRsingles-and-doubles CI(MRSDCI)wave fun tion.
The s attering al ulations ne essary to obtain the potential energy
urvesabovetheionarefarmoretime onsumingthanthestru ture
al- ulations arriedout for those statesbelowtheion. To redu ethe
om-putational time required for the s attering al ulations, but still
main-taining good a ura y for the potential energy urves, natural orbitals
For aCIwavefun tion onstru tedfromorbitals
φ
i
,theele tronden-sity fun tion
ρ
an bewritten asρ =
X
i
X
j
ρ
ij
φ
∗
i
φ
∗
j
,
(2.25)where the oe ients
ρ
ij
are a set of numbers whi h form the densitymatrix.NOsareorbitalsthatredu ethis densitymatrix
ρ
toadiagonalform
ρ =
X
k
b
k
φ
∗
k
φ
∗
k
,
(2.26)and the oe ients
b
k
arein this ase, the o upation numbers of ea horbital. A CI expansion based on su h orbitals will generally have the
fastest onvergen e. A CI al ulationis rst arriedout using theMOs
obtained froma SCF al ulation. The onstru ted densitymatrix
ρ
ij
isthendiagonalizedandtheNOsaredetermined.InthesubsequentMRCI
al ulation, the referen e spa e is built up by using the NOs and this
provides a ompa t representation of theorbitals thatis well suited for
thefollowings attering al ulations.AMRCI al ulationusingtheNOs
instead of the arbitrary basis set will give a wave fun tion onsisting
only of those ongurations built up from natural orbitals with large
o upation number.
2.1.3
Scattering calculations
S attering events an o urbetween dierent kinds of parti les. In this
thesis,Ihave onsidereds attering betweenan ele tronandamole ular
ionin theele tron re ombinationstudies and also s attering between a
ation and an anion in the ase of mutual neutralization. The ele tron
s attering al ulations have been performed in order to determine the
parameters of the resonant states above the ion. These parameters are
theenergy positionand theautoionizationwidthat a xedinternu lear
distan e
R
for a parti ular state. In studying the s attering of anele -tronand mole ular ion,the omplex Kohn variational method hasbeen
used [26, 27℄ and this method is outlined later in this se tion. For the
mutualneutralization pro ess,the potential energy urveshave already
been determined and the dynami s on these urvesis investigated with
nu lear s attering al ulations. The method used in performing these
al ulationsis des ribed furtherinthe se tion onRea tion Dynami s.
General
Consider the s attering pro esses between an in oming ele tron and a
mole ulartargetshowns hemati allyinFig.(2.2).Oneofthedi ulties
ins attering al ulations [28℄is how to handlenon-spheri ally
spheri ally symmetri intera tion potential
V (r)
, wherer
now denotestheradialdistan ebetweentheele tronandthetarget. Theentry
han-nel onsists of the ion in its ground vibrational level and an in oming
ele tronwithwavenumber
k
travelingalongthe ollisionaxis.Oneexam-pleexit hannel anbetheioninitsgroundvibrationora rovibrational
ex itedlevelandtheele trontravelingwithadierentwavenumberand
inadire tion
k
′
.Atlargeinternu leardistan es,thes atteredwave
fun -tionshould onsistofan in omingplanewave intheentran e hannel
j
andoutgoing spheri alwavesintheenergeti allyavailableexit hannels
i
. If the in oming plane wave is taken as traveling along the ollisionaxis(thez-axis)the boundary onditionfors attering from hannel
j
tohannel
i
an be written asχ
d
i
(r, θ) ∼
r→∞
δ
ij
e
ik
j
z
+
f
ij
(θ)e
ik
i
r
r
,
(2.27)where the azimuthal angle
ϕ
dependen e has been omitted due to theylindri alsymmetryaroundthe ollisionaxis.Thefun tion
f
ij
(θ)
isthes attering amplitude and
k
i
is the asymptoti wave number. The rossse tion for therea tion is obtained by integrating the dierential ross
se tion, given by the absolute square of the s attering amplitude, over
allsolidangles
σ
ij
=
2πk
i
k
j
Z
π
0
|f
ij
(θ)|
2
sin θ
dθ .
(2.28)To determine the s attering amplitudes,
f
ij
(θ)
,the asymptoti form ofthes atteredwavefun tion,Eq.(2.27),isexpandedintermsofLegendre
polynomials,
P
l
(cos θ)
,andradialwavefun tions. Theradial wavefun -tions in the asymptoti limit an be written as a linear ombination of
anin oming wave and an outgoing wave multiplied byas attering
ma-trixelement,
S
o
ij
,whi h measures the responseofthe target.When thisexpressionis inserted into the partialwave expansion,and thenal
thefollowing expression for the s attering amplitude
f
ij
(θ)
isobtainedf
ij
(θ) =
i
2(k
i
k
j
)
1/2
∞
X
l=0
(2l + 1) δ
ij
− S
ij,l
o
P
l
(cos θ) .
(2.29)Thesupers ript
o
onthes atteringmatrixelementS
o
ij,l
indi atesthat en-ergeti allyopen hannels aretreated.Inserting theexpressionforf
ij
(θ)
into thatfor the rossse tion, Eq.(2.28),yieldsσ
ij
=
∞
X
l=0
σ
ij,l
,
(2.30) whereσ
ij,l
=
π
k
2
j
(2l + 1)
δ
ij
− S
ij,l
o
2
.
(2.31)Consider a totally elasti ollision in one ofthe hannels
i
. The ross se tionfor this hannel,σ
e
= σ
ii,l
,be omesσ
e
=
π
k
2
∞
X
l=0
(2l + 1)|1 − S
l
|
2
,
(2.32)where
S
l
= S
ij,l
is thes attering, orS
,matrix,thequantity1 − S
l
= T
l
isthetransition, orT
,matrix.Attheresonantenergy,
E
res
,theele tron anbetemporarily apturedby the ioninto a nearly bound state giving rise to a sharp variation in
the rossse tion,i.e. aresonan e. The termresonan e iswidelyusedin
theliterature, butunfortunately,oftenwithdierentmeanings. Sowhat
isit? Supposethat the ele tron is aptured, or temporarily trapped by
theion,at someenergy
E
res
.Atthisenergy theelasti rossse tionwillhave a pole in the
S
(orT
) matrix and, therefore, give rise to a sharpvariation inthe rossse tion lose to the resonant energy
E
res
.Anon-redundant poleinthe
S
(orT
)matrix isalsoazeroof theJost fun tionf
l
[28℄sin eS
l
(k) =
f
l
(−k)
f
l
(k)
.
(2.33)Furthermore,azerooftheJostfun tion,orapoleinthe
S
orT
matrix,also orrespondsto apoint inthe omplex
k
-plane.Three dierent kindofstatesexistdependingonwhere the point
¯
k
islo atedinthe omplexk
-plane:bound state,
k = iκ
¯
, withκ > 0
virtual state,
¯
k = −iγ
, withγ > 0
resonant state,
¯
Sin e
E
isproportionaltok
2
,aresonantstatewillhave omplexenergy
with
E = E
res
− iΓ/2
. IfE
is real, the probability of the state an bewritten as
|ψ(r, t)| =
φ(r)e
−iEt/~
2
= |φ(r)|
2
,
(2.34)but,on theotherhand,if
E
is omplex|ψ(r, t)| =
φ(R)e
−i(E
res
−iΓ/2)t/~
2
= |φ(r)|
2
e
−Γt/~
.
(2.35)Thus,aresonant statehasalifetimeand isnotastationary stateofthe
Hamiltonianfor the system.
TheJost fun tion,
f
l
(k)
,hasa zeroat a resonan e. Nearsu h areso-nan etheJostfun tion anbeexpanded about
¯
k = k
res
− iγ
,a ordingto
f
l
(k) =
df
l
dk
¯
k
(k − ¯k).
(2.36)Provided that
¯
k
is fairly lose to the real axis, Eq. (2.36) is a goodapproximation. Thephaseshift for a xedvalueof
l
,δ
l
(k)
,is given byδ
l
(k) = arg(f
l
) = − arg
df
l
dk
¯
k − arg(k − ¯k)
(2.37)≡ δ
bg,l
+ δ
res
(k)
(2.38)where
δ
bg,l
is the slowly varying ba kground ontribution to the phaseshiftand
δ
res
(k)
isthe resonant partofthephaseshiftthatvaryrapidly.δ
res
(k)
isthe angleshowninFig.(2.3).Whenk
isin reasedsu hthatit movespastthepositionofk
¯
,theresonantpartofthephaseshift,δ
res
(k)
, will in reasefrom0
toπ
.The loserthatk
¯
isto the real axis, themorequi klythis in reasehappens. Ifthephaseshiftinsteadisexpressedas
afun tion of
E
,wehave thatδ(E) ≈ δ
bg,l
+ δ
res
(E),
(2.39)and
sin δ
res
(E) =
Γ/2
[(E − E
res
)
2
+ (Γ/2)
2
]
1/2
.
(2.40)In the spe ial ase where the ba kground phase
δ
bg,l
is zero,σ
l
(E) ∝
sin
2
(δ
l
(E))
,and the expressionfor thepartial rossse tionisgiven byσ
l
(E) ∝
Γ
2
/4
(E − E
res
)
2
+ (Γ/2)
2
.
(2.41)k
-plane.In an a tual ele tron-mole ular ion s attering event, the intera tion
potential,
V (r)
,isnon-spheri al andtheS
(orT
)matrix will have bothl
andm
indi es. Resonant statesappearassharpvariations inthe rossse tion, and even more learly when the eigenphase sum is studied. As
afun tion of energy,the eigenphasesum isgivenby:
δ(E) =
X
l,m
δ
l,m
(R, E),
(2.42)where
R
is the xed internu lear distan e at whi h the al ulation isperformed.Thesuddenshiftoftheeigenphasesumwith
π
attheresonantenergy
E
res
is illustrated in Fig. (2.4), whi h shows the results of anele trons attering al ulationfor
e
−
+
HF+
in1
A
1
symmetryatR = 1.2
a.u..The omplexKohn variationalmethodwasusedinthis al ulation
andthis methodis des ribed inthe nextse tion.
When tting the eigenphase sum inthe neighborhood of a resonan e
tothe following Breit-Wignerform, both theposition,
E
res
(R)
,andtheautoionizationwidth,
Γ(R)
,oftheresonan eatinternu lear distan e,R
, areobtained.δ(E) = δ
res
(E) + δ
bg
(E)
= tan
−1
Γ
2(E − E
R
)
+ a + bE + cE
2
.
(2.43)Here,
δ
res
isthe ontribution fromtheresonan eto the eigenphasesumand
δ
bg
istheba kground ontribution.Toobtainthepotentialenergyoftheresonant state,the ioni potential hasto be added to theresonan e
energy
V
res
(R) = V
ion
(R) + E
res
(R).
(2.44)Complex Kohn variational method
Inthe omplex Kohn variational method [26,27℄, a trial wave fun tion
0
0.05
0.1
0.15
0.2
10
2
Interaction energy (H)
Cross section (cm
2
)
0
0.05
0.1
0.15
0.2
−2
−1
0
1
2
Interaction energy (H)
Eigenphase sum (rad)
(a)
(b)
Figure 2.4: Resonan esfoundin elasti ele trons atteringwith
HF
+
in
1
A
1
symmetry and at an internu lear distan e
R = 1.2 a
0
. (a) shows the rossusingashort-rangespheri allysymmetri potential,
V (r)
,buttheappli- ationto more realisti ases will alsobementioned.Ifthe partialwave
radialS hrödinger equation
Lφ
l
(r) = 0,
(2.45) whereL =
−
1
2
dr
d
2
+
l(l + 1)
r
2
+ V (r) −
k
2
2
,
(2.46)isinserted into the Kohnfun tional,
I
,dened asI[φ
l
(r)] = hφ
l
|L|φ
l
i =
Z
∞
0
φ
l
(r)Lφ
l
(r)dr,
(2.47)then the fun tional
I
is zero if the wave fun tion is the exa t solutionto Eq. (2.45). Conversely, if a trial wave fun tion
φ
t
l
is inserted, thefun tional
I
diersfromzero.Assumethefollowingboundary onditionsfor
φ
t
l
φ
t
l
(0) = 0
(2.48)φ
t
l
(r → ∞) ∼ F
l
(kr) + λG
l
(kr),
(2.49)where
F
l
andG
l
are two linearly independent solutions of the radialS hrödinger equationwith
V (r) = Z/r
.Therst onditionrequiresthatthetrialwavefun tion tobe niteattheoriginwhereasthese ond
on-dition onstrains the ele tron to behave as an outgoing spheri al wave.
The parameter
λ
is a linear oe ient. In order to examine how thefun tional
I
hanges when the exa t solution of the radial S hrödingerequation
φ
l
(r)
isrepla edbythetrialwave fun tion,φ
t
l
(r)
,theresidualbetween thesewave fun tions, dened as
δφ
l
(r) ≡ φ
t
l
(r) − φ
l
(r)
(2.50)withtheboundary onditions
δφ
t
l
(0) = 0
(2.51)δφ
t
l
(r → ∞) ∼ λG
l
(kr)
(2.52)isinserted into the Kohnfun tional. The resultisgiven by:
δI =
Z
∞
0
δφ
l
Lφ
l
dr +
Z
∞
0
φ
l
Lδφ
l
dr +
Z
∞
0
δφ
l
Lδφ
l
dr.
(2.53)UsingEq.(2.45) we anseethattherst terminthis expressioniszero.
Integratingthese ondtermbypartsandmakinguseofEq.(2.48)-(2.50)
gives
δI = −
k
2
W δλ +
Z
∞
0
W = F
l
(r)
d
dr
G
l
(r) − G
l
(r)
d
dr
F
l
(r)
(2.55) andδλ = λ − λ
t
. Here,λ
t
is the variational parameter whereas
λ
istheexa t value. Eq.(2.54) isknown astheKato identity.Thisequation
providesastationaryprin ipleforapproximating
λ
sin etheexa tvaluesof
λ
an beobtained intermsoftheapproximate valueofthefun tionalI[φ
t
l
]
and atermwhi hisse ondorderintheerrorofthewave fun tion.Thestationaryprin iple of approximating
λ
is givenbyλ
s
= λ
t
+
2
kW
Z
∞
0
δφ
t
l
(r)Lδφ
t
l
(r)dr.
(2.56)The trial wave fun tion should des ribe the ele tron well in both the
intera tion region andin theasymptoti region and,for thatreason,
φ
t
l
is hosento beφ
t
l
(r) = f
l
(r) + λ
t
g
l
(r) +
n
X
i=1
c
i
φ
i
,
(2.57)where
φ
i
aresquareintegrable fun tionsandf
l
(r → ∞) ∼ F
l
(kr)
(2.58)g
l
(r → ∞) ∼ G
l
(kr).
(2.59)The oe ients,
c
i
andλ
t
aredetermined fromthefollowingvariational
onditions
∂λ
s
∂c
i
=
∂λ
s
∂λ
t
= 0.
(2.60)Insertingthetrialwave fun tion given byEq.(2.57) intoEq. (2.56) and
takingthederivativewithrespe tto
c
i
givesZ
∞
0
φ
i
Lφ
t
l
dr = 0, i = 1, ..., n.
(2.61) Similarly,withλ
t
,we obtainZ
∞
0
g
l
Lφ
t
l
dr = 0, i = 1, ..., n.
(2.62)By olle tingthebasisfun tions
g
l
(r)
andφ
i
(r)
intoasingleset{φ
i
}, i =
0, ..., n
, withφ
0
≡ g
l
,and denoting the linear parameters{λ
t
, c
1
, ..., c
n
}
bytheve tor
c
,Eq.(2.61)andEq.(2.62) anbeexpressedina ompa t formwhere
M
isa matrix withelementsM
ij
=
Z
∞
0
φ
i
Lφ
j
dr, i, j = 0, ..., n
(2.64)and
s
is ave tor withelementss
i
=
Z
∞
0
φ
i
Lf
l
dr, i = 0, ..., n.
(2.65)The stationary value of
λ
s
is obtained by inserting Eq. (2.63) intoEq.(2.56)whi h gives
λ
s
=
2
kW
Z
∞
0
f
l
Lf
l
dr − sM
−1
s
.
(2.66)Finally,if
g
l
(r)
is hosento be theoutgoing fun tionh
+
l
(r)
,dened ash
+
l
(r) = i[F
l
(kr) − iG
l
(kr)]
1
√
k
,
(2.67)thentheWronskian redu es to
W = −1/k
andEq. (2.66) redu es to anexpressionfor the
T
-matrix,λ
s
= T
s
l
= e
iδ
l
sin δ
l
[28℄T
l
s
= −2
Z
∞
0
f
l
Lf
l
dr − sM
−1
s
.
(2.68)The matrix
M
is now omplex symmetri and its inverse is generallynon-singular at realenergies [29℄,hen ethename omplexKohn.
To use the omplex Kohn variational prin iple for ele tron-mole ular
ions attering,forwhi htheintera tionpotentialisnon-spheri al,amore
appropriatetrialwavefun tionhastobe hosen.Thistrialwavefun tion
mustbe exibleenough todes ribethe behaviorbothintheintera tion
region andalso asymptoti ally.
For this reason, the ele troni trial wave fun tion of the (
N + 1
)-ele trons attering systemis expandedas[26℄
Ψ
=
X
Γ
A[Φ
Γ
(x
1
...x
N
)F
Γ
(x
N +1
)] +
X
µ
d
µ
Θ
µ
(x
1
...x
N +1
).
(2.69)Therstsuminthisexpressionis alledtheP-spa e portionofthewave
fun tion.The sumrunsovertheenergeti ally opentarget states, where
Φ
Γ
are the N-ele tron target eigenstates andx
i
the spa e-spinoor-dinates. The operator
A
antisymmetrizes the oordinates of the targetands atteredele trons.Inthese ondsum, the
Θ
µ
aresquare-integrableN + 1
onguration state fun tions(CSFs). Thesefun tionsareusedtodes ribe short-range orrelations and theee ts of losed hannels and
fun tion.
Two lasses of terms are in luded in Q-spa e. The rst lass of
fun -tionsisthesetof all(
N + 1
)-ele tronCSFsthat anbeformedfromthea tive spa e of target orbitals. They will relax any onstraints implied
by the strong orthogonality between the s attering fun tions,
F
Γ
, andthetarget orbitals and are alledpenetration terms.A se ond lassof
terms, alledCIrelaxationterms, analsobein luded.Inthis lasswe
have all losed hannels that an be formed from the set of target
on-gurations used to build P-spa e. These CI relaxation terms, together
with the penetration terms, onstitute the orrelation part of the trial
wave fun tion.
Whenthe trialwave fun tion is insertedinto the fun tional, and this
fun tionalisminimizedtoobtaintheunknown oe ients,the
T
matrixand thus the eigenphase sum, given by Eq. (2.42), will ontain both
l
and
m
indi es. Thekind of behaviordisplayed inFig.(2.4) an thenbeobtained forthe rossse tion andtheeigenphase sum.
2.1.4
Diabatization
After the adiabati potential energy urves have been obtained, below
theionbystru ture al ulationandabovetheionbyele tron-s attering
al ulations, these urvesare thendiabatized. As mentioned earlier,we
arerestri ted to dierent typesof quasidiabatizations sin e inthe
rea -tions we arestudying thereare aninnite numberof intera ting states.
Themethods thathave been usedfor thepurpose ofquasidiabatization
will be dis ussed next. However, before this dis ussion, Iwill briey
in-trodu e mole ular symmetry sin e stru tureand s attering al ulations
inmost asesare arriedoutinasymmetrythatisnotthetruesymmetry
ofthemole ule.
Restriction of molecular symmetry in the calculations
Thesymmetryofmole ulesisdes ribed intermsofelementsand
opera-tions.Elementsaregeometri entities su hasaxes,planesand pointsin
spa eusedtodenesymmetryoperations.Operationsinvolvethespatial
re-arrangement of atoms in a mole ule by rotation about an axis (
C
n
),by ree tion through a plane (
σ
), or by inversion through a point (i
).A rotation, ree tion or inversion operation will be alled a symmetry
operation if, and only if, the new spatial arrangement of the atoms in
themole ule isindistinguishable fromtheoriginal arrangement.
The point group, or symmetry group, is the name of a a olle tion
ofsymmetryelementspossessed byamole ule.Ea h ommon olle tion