ebook img

The adiabatic strictly-correlated-electrons functional: kernel and exact properties PDF

1.6 MB·
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview The adiabatic strictly-correlated-electrons functional: kernel and exact properties

Theadiabaticstrictly-correlated-electronsfunctional: kernelandexactproperties GiovannaLani,1 SimonediMarino,2 AugustoGerolin,3 RobertvanLeeuwen,4 andPaolaGori-Giorgi1 1Department of Theoretical Chemistry and Amsterdam Center for Multiscale Modeling, FEW, Vrije Universiteit, De Boelelaan 1083, 1081HV Amsterdam, The Netherlands 2LaboratoiredeMathe´matiquesd’Orsay,Univ. Paris-Sud,CNRS,Universite´ Paris-Saclay,91405Orsay,France. 3Dipartimento di Matematica, Univerisita´ di Pisa, Largo B. Pontecorvo, 56126 Pisa, Italy 4Department of Physics, Nanoscience Center, University of Jyva¨skyla¨, 40014 Jyva¨skyla¨, Finland; European Theoretical Spectroscopy Facility (ETSF) (Dated:January18,2016) Weinvestigateanumberofformalpropertiesoftheadiabaticstrictly-correlatedelectrons(SCE)functional, relevantfortime-dependentpotentialsandforkernelsinlinearresponsetime-dependentdensityfunctionalthe- 6 ory. Amongtheformer, wefocusonthecompliancetoconstraintsofexactmany-bodytheories, suchasthe 1 generalisedtranslationalinvarianceandthezero-forcetheorem.Withinthelatter,wederiveananalyticalexpres- 0 sionfortheadiabaticSCEHartreeexchange-correlationkernelinonedimensionalsystems,andwecomputeit 2 numericallyforavarietyofmodeldensities. Weanalysethenon-localfeaturesofthiskernel,particularlythe n onesthatarerelevantintacklingproblemswherekernelsderivedfromlocalorsemi-localfunctionalsareknown a tofail. J 5 1 I. INTRODUCTION In Eq. (1), vH([n],r,t) is the usual Hartree potential computed with the time-dependent density n(r,t), and ] vxc([Ψ0,Φ0,n];r,t) is the exchange-correlation (xc) poten- l e While a considerable amount of work on the strictly- tial, depending also on the initial states Ψ0 and Φ0. Trad- - correlated-electrons (SCE) formalism1–3 within the frame- ing the many-body TDSE for the one-particle TDKS equa- r t workofgroundstateKohn-Sham(KS)densityfunctionalthe- tionshasapricetopay,thatisthetime-dependentexchange- s ory (DFT) has been carried out,4–8 the study of its perfor- correlation potential v ([n,Ψ ,Φ ];r,t) of TDDFT is an . xc 0 0 at mancesinthetimedomainisjuststarting.9,10 Theaimofthis evenmorecomplexobjectthanthevxc forgroundstateDFT, m work is to begin a systematic investigation of the SCE func- asitisafunctionalofthedensityatalltimest(cid:48) ≤ tand,ad- tional in the context of time dependent problems, in order to ditionally, oftheinitialstateofboththeinteractingandnon- - d understand its fundamental aspects and its potential in tack- interacting systems. However, whenever the initial state for n ling challenging problems for the standard approximations the evolution problem described by the TDKS equations is o employedintime-dependent(TD)DFT. chosen to be the ground state of the system, then the func- c tionaldependenceofthexcpotentialisontheelectronicden- [ Wewillhencefocusonthosephysicalsituationsdescribed sityalone: sincethisscenariooccursnaturallyinmanyprob- 1 by an explicitly time-dependent Hamiltonian, and whose lems of interest, it doesn’t pose actual limitations and thus it v dynamics is described by the time-dependent Schro¨dinger isoftenadoptedinpracticalapplications. 7 equation (TDSE). Due to the existence of a time-dependent 7 density-potential mapping11–14 for interacting and non- Similarly to ground state DFT, in order to make use of 9 interacting systems, a time-dependent Kohn-Sham approach Eq. (1) one needs approximations for the exchange correla- 3 0 can be rigorously set up and employed to study the dynam- tion potential vxc. A first drastic approximation, which is . ics of quantum systems at a manageable computational cost. usedinthelargemajorityofcasesinTDDFT,istheso-called 1 Choosingtheinitialnon-interactingwavefunctiontobeasin- adiabatic approximation, obtained by inserting in a ground- 60 gle Slater determinant of some spin orbitals ψj(x,t0), one stateapproximatevxc([n];r,t)theinstantaneousdensity, ig- 1 canreducetheTDSEtoasetofsingle-orbitalequations, the noring dependence on the density at earlier times. This ap- : time-dependent Kohn-Sham (TDKS) equations, of the form proximation has a very specific range of validity – infinitely v (inHartreeatomicunitsusedthroughout): slowly varying perturbations, such that the system is always i X in its ground state – but it is very often employed outside ar i∂tψj(x,t)=(cid:16)− 12∇2+vext(r,t)+vH([n],r,t) (1) idte,pwenitdhinrgesuolntsththeatnactaunrevaorfythfreomprovbelreymsaatdisdfraecstsoerdy.toInpoceorr-, (cid:17) tain cases, it is still difficult to disentangle the errors due +v ([Ψ ,Φ ,n];r,t) ψ (x,t), xc 0 0 j to the adiabatic approximation and the errors due to the ap- with Ψ and Φ initial states of the true interacting and of proximation for the ground-state exchange correlation po- 0 0 tential, but considerable progress has been made in recent the non-interacting KS system. The time-dependent density years, by analysing, whenpossible, the “adiabaticallyexact” is thus computed in the familiar way (for simplicity in this potential.15–17Inothercases,itisinsteadwellestablishedthat introductionweconsiderclosed-shellsystems)as: neglecting all “memory effects” in v (or equivalently fre- xc N/2 quency dependence in the so called xc kernel, Fxc of linear (cid:88) n(r,t)=2 |ψ (r,t)|2. (2) response TDDFT), does not allow TDDFT to describe ex- j citations with a predominantly double character.18–22 In the j=1 2 TDDFTframework,adiabaticityisthusequivalenttolocality tations. Finallywewillgiveourconclusionsandperspectives intime. Themostcommonapproximationsusedtobuildthe forfuturework. adiabaticv (andF inthelinearresponsecase)inTDDFT, xc xc arelocalandsemi-localfunctionals,whichthusaddtolocal- ityintimelocalityinspaceaswell. Inthesecases,thetime- dependentkernelFxc(r,r(cid:48),t,t(cid:48))isapproximatedas II. REVIEWOFTHESCEFORMALISM δ2Eappr[n] F ([n];rt,r(cid:48)t(cid:48))= xc , (3) The SCE formalism can be put in the DFT context start- xc δn(r,t)δn(r(cid:48),t(cid:48)) ingwiththegeneralizationoftheHohenberg-Kohnfunctional F[n]toscaledinteractions: whereEappr[n]isevaluatedattheinstantaneousdensityand xc itisoftenalocalorsemi-localapproximatefunctional,which F [n]= min(cid:104)Ψ|Tˆ+λVˆ |Ψ(cid:105) (4) makesthekerneldifferentfromzeroonlyon(orverycloseto) λ ee Ψ→n thediagonalr = r(cid:48). Wehavealreadyhintedattheshortcom- ings of the locality in time in this introduction, but also the where Tˆ and Vˆ are the familiar kinetic and two-body in- ee localityinspacehasseriouslimitations,anotoriousexample teraction operators, while λ is a parameter varying continu- beingthedescriptionofexcitationswithalong-rangecharge ously from 0 to ∞, yielding different scenarios: F [n] = λ=0 transfer (CT) character23,24. In the case of closed-shell frag- T [n] corresponds to the non interacting or Kohn-Sham sys- s ments, the introduction of a considerable portion of Hartree- tem, F [n] corresponds to the real physical system, while λ=1 Fock exchange (often introduced at long-range only through F [n]definesthestrong-couplinglimit,1,2capturedbythe λ=∞ range-separation) is able to fix the CT problem in linear re- strictly-correlated-electronfunctional sponseTDDFT.25–29However,thissolutiondoesnotworkfor theverychallengingcaseofhomolyticbondbreakingexcita- VSCE[n]≡ min(cid:104)Ψ|Vˆ |Ψ(cid:105). (5) ee ee tions, the prototypical example being the lowest excited sin- Ψ→n gletstate1Σ+u oftheH2molecule.30,31Inthiscase,thekernel TheworkinghypothesistobuildtheminimizerofEq.(5)fora shoulddivergeinordertocompensatethefactthatthisexcita- givendensityisthatinthislimitthemany-bodywavefunction tionintheKSsystemgoestozeroasthebondisbroken.30,31 collapsesintoa3-dimensionalsubspaceofthefullconfigura- In this context, we will show that, at least in a model one- tionspace, dimensional case, the adiabatic SCE (ASCE) kernel shows a verypromisingnon-localdivergingbehavior. 1 (cid:88)(cid:90) n(r) In order to construct approximations both for potentials and |ΨSCE(r1,...,rN)|2 = N! dr N δ(r1−f℘(1)(r)) kernelsinTDDFT,onecanbeguidedbytryingtosatisfyexact ℘ propertiesandconstraintsofmany-bodytheories. Inaseries ×δ(r −f (r))···δ(r −f (r)), (6) 2 ℘(2) N ℘(N) of works32–34 Dobson and Vignale devised a number of con- straints(namedtheoremsafterwards)thatthetime-dependent where℘denotesapermutationof1,...,N,suchthatn(r)= v shouldcomplyto, inordertoavoidunphysicalresultsor N(cid:82) |Ψ (r,r ,...,r )|2dr ···dr . The functional is xc SCE 2 N 2 N contradictionsinthetheory. Fromtheiranalysis, itappeared then specified in terms of the so-called co-motion functions for the first time that the interplay between non locality in f ([n];r)thatdeterminethesetinwhich|Ψ |2 (cid:54)=0, i SCE space and non locality in time is a delicate issue in TDDFT and this fact needs to be kept in mind when looking for ap- 1(cid:90) (cid:88)N 1 VSCE[n]= n(r) dr. (7) proximations, making this task much more challenging than ee 2 |r−f ([n];r)| i ingroundstateDFT.Itisthusnaturaltoaskwhetherahighly i=2 non-localfunctionalsuchasSCEcansatisfytheseexactcon- ThefunctionalderivativeofVSCE[n]definestheSCEpoten- ditionswhenemployedintheadiabaticapproximation. ee tial: After briefly reviewing in Sec. II the basics ideas of the δVSCE[n] SCE formalism, we will show in Sec. III how the SCE po- vSCE([n];r)= ee , (8) δn(r) tential satisfies exact properties of many-body theories, such asthezero-forcetheoremandthegeneralizedtranslationalin- which can be computed via a rigorous and physically trans- variance. Ouranalysiswillalsoshowhow,whilenon-locality parentshortcut2astherepulsionfeltbyanelectroninrdueto in time and non-locality in space have to go hand in hand, theotherN −1electronsatpositionsr =f ([n];r), non-localityinspaceandlocalityintimecancoexistwithout i i violatingtheabovementionedproperties. InSec.IVwewill N derive an analytical expression for the SCE kernel for one- ∇vSCE([n];r)=−(cid:88) r−fi([n];r) . (9) dimensionalsystems,andthencomputeitnumericallyforvar- |r−fi([n];r)|3 i=2 iousdensityprofiles. Wewillcompletethesectionwithadis- cussiononsomegeneralfeaturesofthekernel,pinpointingat All the f [n](r), whose physical meaning is to give the posi- i thosewhicharisefromitshighlynon-localnatureandwhich tions of all the others N − 1 electrons once the position of couldbepromisingforthedescriptionofbond-breakingexci- areferenceelectronhasbeenfixedinr,satisfythefollowing 3 non-lineardifferentialequation: dependent density, the time-dependent xc potential is rigidly translatedbythesamequantity. Wewillrefertothisproperty n(fi([n];r))dfi([n];r)=n(r)dr i=2,..,N −1 (10) as generalized translational invariance (GTI), since it holds alsoforcoordinatesframeswhichareacceleratedwithrespect which shows their non-local dependence on n(r). Further- totheoriginalone.43 more the co-motion function obey (cyclic) group properties2 ThusingeneraltheGTIcanbeformalizedasfollows: given whichensurethattheelectronsareindistinguishable. an arbitrary (be or not time-dependent) shift of the density Intherecentyears,ithasbeenrealizedthattheproblemde- R(t): finedbytheminimization(5)isequivalenttoanoptimaltrans- portproblemwithCoulombcost.35,36 Sincethen,theoptimal n(cid:48)(r,t)=n(r−R(t),t) (11) transport community has been able to prove several rigorous results. Inparticular,theSCEstate(6)hasbeenproventobe thexc potentialassociated withthis densityhas totransform the true minimizer for any number of particles N in one di- accordinglyto: mensional(1D)systems37andinanydimensionforN =2.35 Formoregeneralcases,ithasbeenshownthattheminimizer vxc([n(cid:48)];r,t)=vxc([n];r−R(t),t) (12) mightnotbealwaysoftheSCEform38. Eveninthosecases, however,SCE-likesolutionsseemtobeabletogoveryclose tothetrueminimum,39 andinseveralcasesitisstillpossible A. PropertiesoftheadiabaticSCEpotential toproveEq.(9).39 Inthelow-densitylimit(orstrong-couplinglimit)theexact WewillnowshowexplicitlyshowthattheASCEcomplies Hartree andexchange-correlation (Hxc)energy functionalof KS DFT tends asymptotically to VSCE[n],36,40 Thus, in the tothisrequirement. ee followingwedenotevSCE([n];r)ofEq.(9)asvSCE([n];r),to We begin by observing that in the SCE limit, upon the shift Hxc ofthedensity,alltherelativedistancesbetweentheelectrons stressthatthispotentialisthestrong-couplingapproximation tothestandardHxcpotentialofKSDFT.4,40 havestilltobethesame, thustheco-motionfunctionstrans- formas: Now that the basics of the SCE formalism at the ground state level have been reviewed, we can move to the time- f ([n(cid:48)];r)=f ([n];r−R(t))+R(t). (13) dependentdomain. i i In the one dimensional case, where the co-motion functions can be expressed in terms of a simple one-dimensional inte- III. EXACTPROPERTIESFROMMANY-BODY gral,onecanshowexplicitlythattheaboverelationholds,see THEORIES App. A for details. Substituting the transformed co-motion functionsintotheexpressionfortheSCEpotentialgives: Since the success of TDDFT relies heavily on the avail- N abilityandthequalityoftheapproximationsforvxc([n];r,t) ∇vSCE([n(cid:48)];r,t)=−(cid:88) r−fi([n];r−R(t))−R(t) and for the linear response exchange-correlation kernel Hxc |r−fi([n];r−R(t))−R(t)|3 i=2 F ([n],r,r(cid:48),t,t(cid:48)), there have been intense research efforts xc =∇vSCE([n];r−R(t),t). (14) towards better approximations. As already mentioned in the Hxc introduction,awaytoguidesuchapproximationsistoresort IntegrationandsubtractionoftheHartreepotential(whichsat- to the compliance to exact constraints from many-body the- isfytheGTIstraightforwardly)yields: ories, similarly to what has been done extensively already in ground state DFT. A first exact condition is given by scal- vSCE([n(cid:48)];r,t)=vSCE([n];r−R(t),t) (15) xc xc ing relations,41,42 a second one by a sum rule for the time- dependent exchange-correlation energy41 and just like in the whichistherelationwewantedtoprove. static case, the time-dependent xc potential should be self- A second important constraint is that the xc potential can- interactionfree. notexertanetexternalforceonthesystem,whichisnothing In addition to the constraints enumerated above, a very im- elsethanthecompliancetoNewton’sthirdlawofmotion. In portant condition on approximate xc potentials is that they DFTthispropertygoesunderthenameofzeroforcetheorem shouldbeGalileaninvariant,asaconsequenceofthefactthat (ZFT)andin33 itwasshownhowitisautomaticallysatisfied the TDSE itself exhibits this symmetry. This condition was fortranslationallyinvariantxcpotentials. Onemaythinkthat firstinvestigatedbyVignale,33asageneralizationofanearlier this is a trivial requirement to be satisfied, but in practice it work by Dobson32 on the so called harmonic potential theo- isn’t. Forexamplein44 itwasdemonstratednumericallythat rem (HPT), which states that upon the application of a time- computingthedipolemomentofsmallNa andNa+clusters, 5 9 dependent field to a many-body system confined by an har- via the exact exchange Krieger-Li-Iafrate approximation to monicpotential,itstime-dependentdensityisrigidlyshifted. v ,yieldedanincreasedamplitudeinthedipoleoscillations, xc In33 it was demonstrated that the HPT is automatically sat- mostlikelyduetospuriousinternalforcesappearingasacon- isfiedwheneverthetime-dependentxcpotentialobeysapre- sequence of the violation of the ZFT. The ZFT not only has ciseconstraint,thatisuponarigidshiftofthesystem’stime- implicationsfortheapproximationstothetime-dependentxc 4 potential,butalsoonanotherkeyquantityofTDDFT,namely B. PropertiesoftheadiabaticSCEkernel the exchange correlation kernel. In34 Vignale showed how a frequency dependent (thus non local in time) F cannot be xc Let’snowturntotheASCEkernel, local in space, in order to satisfy the ZFT. A notable exam- ple of a kernel which violates the ZFT and the HPT too, is δ2VSCE[n] theGross-KohnFxc45 whichindeedisfrequencydependent, FHAxScCE([n];rt,r(cid:48)t(cid:48))= δn(r,te)en(r(cid:48),t(cid:48))δ(t−t(cid:48)), (21) butlocalinspace,asitisbasedonthehomogeneouselectron gas. This peculiar issue in TDDFT is commonly known as OnceagainweresorttoanexpansionforthedensityinR(t), ultra non-locality problem and makes particularly challeng- thatisn(r−R(t),t)≈n(r,t)−R(t)·∇n(r,t),combining ingtheconstructionofapproximatefrequencydependentker- thiswithEq.(15)andinvokingthearbitrarinessofR(t)and nels. AdiabaticFxc, derivedfromfullylocalfunctionals, do thedefinitionofASCExckernel,weobtain: not violate the ZFT. It is legitimate to ask if an adiabatic but (cid:90) highly non local functional like the ASCE, does violate the dr(cid:48)FASCE([n];rt,r(cid:48)t(cid:48))∇n(r(cid:48))=δ(t−t(cid:48))∇vSCE([n];r,t), ZFT.Strictlyspeakingwealreadyknowthatitdoesn’t,sinceit Hxc Hxc respectstheGTI,butinthefollowingwewillexplicitlyshow whichshowsthattheASCEkernelindeedsatisfiestheZFTin thatwhilenonlocalityintimerequiresnonlocalityinspace, theconverseisnottrue. thelinearresponseregime. Let’s consider once again a shift in the density: n(cid:48)(r) = n(r−R). Observing that the generalized HK energy func- tionalFλ[n]istranslationallyinvariant(sincebothTˆandVˆee C. Propertiesoftheco-motionfunctions are)onehas: Atthispointitseemsnaturaltoalsoinvestigatesomeprop- F [n]=F [n(cid:48)]. (16) λ λ erties of the co-motion functions. Combining again the ex- pansionforthedensityofEq.(17)withEq.(13)oneobtains: ExpansionofthedensityinpowersofRgives: n(cid:48)(r)=n(r−R)=n(r)−R·∇n(r)+O(R2), (17) (cid:90) dr(cid:48)δfi,α([n];r) ∂ n(r(cid:48))= ∂ f ([n];r)−δ (22) δn(r(cid:48)) ∂r(cid:48) ∂r i,α αβ β β andexpandingbothsidesofEq.(16)yields: where α,β run over Cartesian indices x,y,z. Eq. (22) is a (cid:90) δF 0= dr λ (−R·∇n(r)) (18) sumrulethatcanbewrittenalsoforadiabatictime-dependent δn(r) co-motion functions and the static density and may be em- ployed as constraint to devise approximate co-motion func- whichisvalidforanyarbitraryshiftR. tions. Furthermoreitcanbeusedtoverify(particularlyinthe ThecaseF = VSCE correspondstotheSCEfunctional, λ=∞ ee easier one-dimensional case) the functional variation of the hence: co-motionfunctionswithrespecttothedensity. (cid:90) 0= drvSCE([n];r)∇n(r) (cid:90) =− dr∇vSCE([n];r)n(r) (19) IV. SCEHARTREE-EXCHANGECORRELATION KERNELFORONE-DIMENSIONALSYSTEMS which shows that the SCE potential does indeed satisfy the ZFTforstaticdensities.Additionally,sincethedifferentiation In the one-dimensional case with convex repulsive inter- above is completely general and holds for any density, even particleinteractionw(|x|), theSCEsolution1 isknowntobe time-dependentones,onehas: exactforanynumberofelectronsN,37 andcanbeexpressed inarathersimpleformintermsofthefunctionN ([n];x), (cid:90) e 0= dr∇vSCE([n];r,t)n(r,t), (20) (cid:90) x N ([n];x)= n(y)dy, (23) e −∞ whichshowsthattheASCExcpotentialsatisfiestheZFTfor time-dependentdensitiesaswell. andofitsinverseN−1([n];x): e f ([n];x)=f+([n];x)θ(x−a [n])+f−([n];x)θ(a [n]−x) i i k i k whereθ(x)istheusualHeavisidestepfunction,and f+([n];x)=N−1([n];N ([n];x)+i−1) (24) i e e f−([n];x)=N−1([n];N ([n];x)+i−1−N), i e e witha [n]=N−1([n];k),andk =N +1−i. k e 5 InthiscasetheSCEpotentialissimplygivenby thekernelisgivenby N (cid:90) ∞ FSCE([n];x,x(cid:48))=(G(x(cid:48))−G(x)+G(0))θ(x(cid:48)−f(x)) (cid:88) Hxc vHSCxcE([n];x)=− w(cid:48)(|y−fi([n];y)|)× for x>0, x(cid:48) <0. (32) i=2 x sgn(y−fi([n];y))dy. (25) TheresultingSCEFSCE([n];x,x(cid:48))forthiscaseisplottedin Hxc the first panel of Fig.1: as it is evident from Eq. (29), the TheSCEkernelisthenequaltothevariationoftheSCEpo- kernel has in the first and third quadrants (x,x(cid:48) > 0 and tentialwithrespecttotheelectrondensity, x,x(cid:48) < 0) the same value as along the diagonal (x = x(cid:48)), while in the second and fourth quadrants (Eq. (32)) the ker- δvSCE([n];x) FSCE([n];x,x(cid:48))= Hxc , (26) nelisdifferentfromzeroonlyintheregiondelimitedbythe Hxc δn(x(cid:48)) x,x(cid:48) axesandtheco-motionfunctionx(cid:48) =f(x). Thebehav- ior of an adiabatic kernel local in space, such as the ALDA, whichcanbecarriedout(thedetailsofthederivationaregiven isinsteadradicallydifferent: theFALDA(n];x,x(cid:48))hasanon- in Appendix B-C), yielding (for densities supported on the xc zero component only along the diagonal, δ(x−x(cid:48)), and the wholerealline)thecompactexpression Hartreecomponent,equaltow(|x−x(cid:48)|),hasamaximumon FSCE([n];x,x(cid:48))=(cid:88)N (cid:90) ∞ w(cid:48)(cid:48)(|y−fi([n];y)|)] (27) thediagonal,decayingas1/|x−x(cid:48)|outsideit. Hxc n(f ([n];y)) i=2 x i ×[θ(y−x(cid:48))−θ(f ([n];y)−x(cid:48))]dy. B. ModelHomonuclearmolecule i Fromthisexpression,itisnotevidentthatEq.(27)satisfiesthe Thesecondtypeofdensityconsideredisamodel2-electron symmetry requirement FSCE([n];x,x(cid:48)) = FSCE([n];x(cid:48),x). Hxc Hxc densitywhichresemblestheoneofahomonuclearmolecule, AnexplicitproofofthissymmetryisgiveninAppendixD. a(cid:16) (cid:17) n(x)= e−a|x−R2|+e−a|x+R2| , (33) 2 A. AnalyticalExample where a = 1 and where R, the distance between the two nuclei, canbeincreasedarbitrarilytosimulatethemolecular bond stretching. For this case we numerically computed the WebeginbyconsideringN =2electronsintheLorentzian SCEkernelfordifferentvaluesofR,toobtaininsightsonhow densityprofile, ahighlynon-localkernelbehavesforaproblemwhichbears 2 1 aresemblancetotheH2 dissociation. TheresultsforR = 3, n(x)= , (28) R = 8 and R = 12 are presented respectively in panels (b), π1+x2 (c)and(d)ofFig.1. forwhichtheco-motionfunctionissimplyf (x) ≡ f(x) = Aside from the peak in the origin, a very interesting fea- 2 −1.FromthegeneralexpressionofEq.(27),weobtaininthe ture displayed by the SCE kernel is the appearance, as R is x firstquadrant increased,oftwoplateaux,eachoccupyingalargesquarere- gion (of size ≈ R×R) of the first and the third quadrants. FSCE([n];x,x(cid:48))=G(−max{x,x(cid:48)}) for x>0, x(cid:48) >0, As in the case of the lorentzian density, the SCE kernel has Hxc (29) alsonon-zerocomponentsinthesecondandfourthquadrants, wherewehavedefinedthefunctionG(x) but they are now much smaller than the ones in the I and III quadrants. (cid:90) x w(cid:48)(cid:48)(|y−f([n];y)|)] The height of the plateaux increases as R increases. A G(x)= dy, (30) n(f([n];y)) closer analysis of the function G(x) defined by Eq. (30) for −∞ the case of the density (33) (see Appendix E) shows that the which in this case, and with e-e interaction w(|x|) = 1 heightandsizeoftheplateauxareapproximatelygivenby |x| (sinceintheSCEwavefunctiontheparticlesnevergetontop of each other, the 1/x divergence at x = 0 in 1D does not 1 FSCE([n];x,x(cid:48))≈ poseanyproblem),isequalto Hxc n(0)(R−1/a)2 1 1 1 1 (cid:40) π 1 x≤0 for (cid:46)|x|(cid:46)R− , and (cid:46)|x(cid:48)|(cid:46)R− , G(x)= 21+x2 (31) a a a a π211++2xx22 x>0. withx,x(cid:48) ≥0 or x,x(cid:48) ≤0 (34) Sinceourdensitysatisfiesn(−x)=n(x),inthiscasetheker- As an example, we show in Fig. 2 the SCE kernel along the nel in the third quadrant (x < 0 and x(cid:48) < 0) is equal to the diagonalforR=8,12and20,multipliedbyn(0)(R−1/a)2, oneinthefirstquadrant.Inthesecondquadrant–andbysym- where n(0) = e−aR/2. The value of the SCE kernel on the metrythefourth,sinceFSCE([n];x,x(cid:48))=FSCE([n];x(cid:48),x)– diagonalalsodefinesthevalueofthekernelinthewholefirst Hxc Hxc 6 andthirdquadrants,seeEq.(29). Thus,weseethattheSCE e-R2(R-1)2FSCE(x,x) Hxc 1.5 R=20 1.0 (a)SCEkernelforaLorentziandensity. 0.5 R=12 R=8 -20 -10 10 20 x Figure2. TheSCEkernelforthe“homonucleardimer”densityof Eq.(33)witha = 1anddifferentinternuclearseparationsRalong thediagonalx = x(cid:48). Thekernelhasbeenmultipliedbyn(0)(R− (b)SCEkernelfora“homonucleardimer” 1/a)2,toshowitsscalingwithR. Thevaluealongthediagonalis densitywithR=3 exactlythesameasthevalueofthekernelinthewholefirstandthird quadrant,seeEq.(29)andFig.1. kerneldevelopsplateauxregionswhoseheightdivergesexpo- nentiallyasRisincreased. Thisdivergenceisverypromisingtocapturebond-breaking excitations, because it makes diverge matrix elements of the kernel between atomic orbitals centered on the same site. Considerthebasicexampleofthelowestexcitedsingletstate 1Σ+oftheH molecule,30,31wherewehaveamatrixelement u 2 (c)SCEkernelfora“homonucleardimer” ofthekind densitywithR=8 (cid:90) (cid:90) dx dx(cid:48)σ (x)σ (x)FSCE([n];x,x(cid:48))σ (x(cid:48))σ (x(cid:48)), g u Hxc g u (35) whereσ = c (φ ±φ ),withφ theatomicorbitals g,u u,g A B A,B centeredinthetwoatoms,andc anormalizationconstant. u,g IntheTDDFTlinearresponseequationsthismatrixelementis multipliedbythecorrespondingKSorbitalenergydifference (cid:15) −(cid:15) , which goes to zero as R → ∞, so that the kernel u g matrix element must diverge in order to keep the excitation finite (as it is in the exact system).30,31 We see that for the termscenteredonthesameatomAappearingin(35)wehave, (d)SCEkernelfora“homonucleardimer” densitywithR=12 forlargeR, Figure 1. The SCE Hartree-exchange-correlation kernel for (cid:90) (cid:90) N = 2 electrons for the Lorentzian density profile of dx dx(cid:48)|φA(x)|2FHSCxcE([n];x,x(cid:48))|φA(x(cid:48))|2 Eq. (28) (a) and for “homonuclear dimer” densities n(x) = (cid:16) (cid:17) 1 1 e−|x+R/2|+e−|x−R/2| , with R = 3 (b), R = 8 (c) and ≈ . (36) 2 n(0)(R−1/a)2 R = 12 (d). The FSCE([n];x,x(cid:48)) displays, in all the four cases, Hxc apeakstructureincorrespondenceoftheverticalasymptotesofthe Equation(36)holdsbecausetheproductoftheatomicorbitals co-motionfunctions. Twosymmetricquasi-plateaux,inthefirstand withthesamecenter, |φ (x)|2|φ (x(cid:48))|2, issignificantlydif- A A thirdquadrants,appearforthestretched“homonucleardimer”den- ferent from zero only in one of the two plateau regions of sities: theybecomegreaterinheightandflatterasthevalueofRis Eq.(34),wherewecanapproximatetheSCEkernelwiththe increased. constantvalue 1 ,whichdivergesexponentiallyas n(0)(R−1/a)2 R→∞. Althoughamorecarefulanalysisisneededtoverify if this divergence is really able to compensate the vanishing 7 oftheKSexcitation(cid:15)u−(cid:15)g comingfromaself-consistentKS AppendixA:Explicitcalculationoftheshiftedco-motion SCE calculation, we see that the FSCE([n];x,x(cid:48)) embodies functionsin1D Hxc the right physics: its very non-local dependence on the den- sitymakesitdivergeintheatomicregion,onlywhenanother Let us consider the negative semi-axis (the positive one distantatomispresent. gives an analogous result). The function N ([n];x) for a e Finally, the height of the peak in the origin can be easily shifteddensityn(cid:48)reads: obtained from the properties of the function G(x) (see Ap- pendixE),anditcanbeshowntobealwaysequalto x x−R(t) (cid:90) (cid:90) (cid:18) R R(cid:19) 2 Ne([n(cid:48)];x)= n(cid:48)(y)dy = n(y−R(t))dy FSCE([n];0,0)=2FSCE [n]; , ≈ . Hxc Hxc 2 2 n(0)(R−1/a)2 −∞ −∞ (37) =N ([n];x−R(t)), (A1) e andtakingitsinverse, V. CONCLUSIONSANDPERSPECTIVES N−1[N ([n(cid:48)];x)]=x e e N−1[N ([n];x−R(t))]+R(t)=x. (A2) InthisworkwehaveexploredtheSCElimitinthecontext e e of time-dependent problems, focusing on the formal proper- Combiningtheaboverelationswehave: tiesoftheadiabaticSCE(ASCE)functional. Wefirstexam- ined some properties of the ASCE time-dependent potential, f ([n(cid:48)];x)=N−1[N ([n(cid:48)];x)+i−1] (A3) i e e inparticularthecompliancetoconstraintsofexactmany-body =N−1[N ([n];x−R(t))+i−1]+R(t) theories, such as the generalized translational invariance and e e =f ([n];x−R(t))+R(t) the zero-force theorem, and showed that the ASCE satisfies i both. Whileitiswellknownthatnon-localityintimerequires whichisthe1DversionofEq.(13). nonlocalityinspace,wehaveshownthattheconverseisnot trueusingtheexampleoftheASCE. Inthesecondhalfofthepaperwederivedananalyticalex- pressionfortheSCEHartreeexchange-correlationkernelfor AppendixB:Functionalvariationofthe1Dco-motionfunctions withrespecttothedensity one-dimensional problems, and we have computed it numer- ically for various density profiles. In particular, we have an- alyzedthecaseofamodelhomonuclear2-electronmolecule LetnbeadensityofameasureinRsuchthat(cid:82) n(x)dx= R asthebondisstretched,findingthattheSCEkerneldisplaysa N. Then, the co-motion functions (or the optimal transport verypromisingdivergingbehaviorthatcouldtackletheprob- maps) are given explicitly by Eqs. (24)-(25), which corre- lemofhomolyticbond-breakingexcitations. spondtothecondition(cid:82)fi(x)n(y)dy ∈{i,−N+i},depend- x In future works we will implement the whole linear re- ingonwhetherx≤a ornot. i sponse TDDFT equations for one-dimensional problems us- Weconsidernε(x)=n(x)+ε(n˜(x)−n(x))andwewant ingtheSCEkernel,analysingifitsdivergingbehaviorisable to determine the corresponding co-motion functions fε. For i toopenthegapinamodelMottinsulator,madeofachainofH every x ∈ R we define x as the point such that f (x ) = ε i ε atoms. Workonbond-breakingexcitationsinrealtimeprop- fε(x),andthenwenoticethat i agationwiththeASCEpotentialhasalsoshownverypromis- ingresultsinthissense,andiscurrentlyinpreparation.10Last (cid:90) fi(xε) (cid:90) fiε(x) but not least, we will use our insights to design approximate n(y)dy− nε(y)dy ∈{−N,0,N}, kernelsbasedontheSCEform. xε x (cid:90) fε(x) (cid:90) fε(x) (cid:90) fε(x) i i i n(y)dy− n(y)−ε (n˜(y)−n(y))dy ACKNOWLEDGMENTS xε x x ∈{−N,0,N}, (cid:90) x (cid:90) fε(x) It is our pleasure to dedicate this paper to Evert Jan i n(y)dy−ε (n˜(y)−n(y))dy ∈{−N,0,N}. Baerends,whohasbeenapioneerinunderstandingandusing xε x linear response TDDFT, and a wonderful mentor to some of (B1) us. WethankAndre´Mirtschink,KlaasGiesbertzandMichael Seidlformanyusefuldiscussionsandforacriticalreadingof SincethislastquantityisforsurelessthanN initsabsolute the manuscript. This work was supported by an IEF fellow- value,thenitmustbeequalto0andso shipfromtheMarieCurieprogram(n.629625)andfromthe European Research Council under H2020/ERC Consolidator (cid:90) x (cid:90) fiε(x) n(y)dy =ε (n˜(y)−n(y))dy; (B2) Grant“corr-DFT”(GrantNo.648932).RvLlikestothankthe AcademyofFinlandforsupport.AGacknowledgesCNPqfor xε x thefinancialsupportofhisPh.D.,whenthisworkwasdone. thisimpliesforsurethatifn>0everywherethenx −xisof ε 8 orderεandtheneitherf −fεisoforderεor(whenx=a ) theregularpart i i i f ∼ ±∞andfε ∼ ∓∞. Inthefirstcaseitisclearthatthe i i right-handsideof(B2)canbeapproximatedby(cid:82)xfi(x)(n˜−n) K (x,x(cid:48))=(cid:88)N (cid:90) ∞w(cid:48)(cid:48)(|y−f ([n];y)|)· losingalowerorderterm;infactthisistruealsointhesecond R i case, using the fact that n˜ −n has null total integral. Now, i=2 x definingthequantity θ(y−x(cid:48))−θ(fi([n];y)−x(cid:48))dy. (C1) n(f (y)) i (cid:90) fi(x) αi(x)= (n˜(y)−n(y))dy (B3) The singular part is more delicate as the classical chain rule x doesnotapplyanymore: wefindthatδf /δnisproportional i andusing(B2), thelastobservation, andthecontinuityofn, to(fi)(cid:48). [Thechainruleforderivativeswhenwehaveastep wecaninferthat functionisnottrivial:justtaketheexampleofg(θ),forwhich thederivativeis(g(1)−g(0))δ andnot(g(cid:48)◦θ)·θ(cid:48) =g(cid:48)(0)δ .] 0 0 α (x) x =x−ε i +o(ε). (B4) In this case, the term coming from the jump, noticing that ε n(x) sgn(a −f±([n];a ))=∓1,hastheform i i i Thismeansthatwehave N (cid:88) (cid:16) fiε(x)−fi(x) = fi(xε)−fi(x) ∼−(f (x))(cid:48)αi(x), (B5) KS(x,x(cid:48))= θ(ai−x)· w(cid:48)(|ai−fi+([n];ai)|)+ ε ε i n(x) i=2 (cid:17) α (a ) w(cid:48)(|a −f−([n];a )|) · i i . (C2) wherethederivativeisadistributionalderivative,whichtakes i i i n(a ) i into account also the jumps. Away from x = a , which is i the only point where the jump can occur, we can make a In particular, whenever n(x) > 0 on the whole real line we simplification, as we have fi(cid:48)(x) = n(x)/n(fi(x)), and so would have f±([n];ai) = ±∞ and this term becomes 0 if (f (x))(cid:48)αi(x) = αi(x) . For the jump we will get an addi- w(cid:48)(∞) = 0whichisthecasefortheCoulombinteractionor i n(x) n(fi(x)) anysituationinwhichtheforce tendsto0 withthedistance. tionaltermintheformofadeltafunction, FordensitiessupportedonthewholeRwehaveonlytheterm α (x) (C1),whichcoincideswithEq.(27). (f+([n];a )−f−([n];a ))δ(x−a )· i . (B6) i i i i i n(x) However, if the support of n is compact, denoting Intheend,noticingthat−αi(x) = (cid:82)Rδn(x(cid:48))·(cid:0)θ(x−x(cid:48))− by n−,n+ the extremes of the support, we would have θ(fi([n];x)−x(cid:48))(cid:1)dx(cid:48),weget f±([n];ai)=n±andthecontributionwouldbenonzero. We canwriteitinaclearerform(usingθ(f+([n];a )−x(cid:48))=0) i i δfi([n];x) =(cid:0)θ(x−x(cid:48))−θ(f ([n];x)−x(cid:48))(cid:1)· δn(x(cid:48)) i N (cid:88) (cid:16) (cid:17) (cid:18) 1 f+([n];a )−f−([n],a ) (cid:19) KS(x,x(cid:48))= θ(ai−x)· w(cid:48)(|ai−n+|)+w(cid:48)(|ai−n−|) · n(f ([n];x) + i in(a )i i ·δ(x−ai) i=2 i i θ(a −x(cid:48)) (B7) i . (C3) n(a ) i AppendixC:FunctionalvariationfortheSCEpotential(kernel) Now we want to use the results of the previous section AppendixD:SymmetryoftheSCEkernelin1D on the variation of the co-motion functions to derive an expression for δvSCE(x)/δn(x(cid:48)), where vSCE(x) is given Hxc Hxc in Eq. (25). We can write FSCE(x,x(cid:48)) = K (x,x(cid:48)) + Hxc R KS(x,x(cid:48)),whereKR andKS are,respectively,thecontribu- The symmetry in x and x(cid:48) of the singular term KS(x,x(cid:48)) tionduetotheregularandsingularpartof δfi(x). Whenthe ofEq.(C3)isobvious,whileforK (x,x(cid:48))ofEq.(C1)thisis δn(x(cid:48)) R co-motionfunctionsdonothavejumps(thatis,wheny (cid:54)=a ) subtler. In order to prove it we compute ∂2 K (x,x(cid:48)): if i ∂x∂x(cid:48) R wecansimplyapplythechainruleandfindanexpressionfor wecanprovethatthisquantityissymmetric,thenthesymme- 9 tryofthewholekernelisautomaticallyproven: AppendixE:PropertiesofthefunctionG(x)forthe homonuclear1Ddensity ∂2 K ([n];x,x(cid:48))= ∂x∂x(cid:48) R Forthe2-electrondensityofEq.(33)witha = 1itiseasy (cid:88)N w(cid:48)(cid:48)(|x−f ([n];x)|)δ(x(cid:48)−x)−δ(x(cid:48)−fi([n];x)) toshowthattheco-motionfunctionsatisfies i n(f ([n];x)) (cid:18) 2 (cid:19) i=2 i f([n];x→0+)=ln(x)−R+ln , (E1) 1+e−R =(cid:88)N w(cid:48)(cid:48)(|x−fi([n];x)|)δ(x(cid:48)−x)− n(f ([n];x)) yielding i i=2 (cid:88)N w(cid:48)(cid:48)(|x−x(cid:48)|)δ(x(cid:48)−fi([n];x)). (D1) n(f([n];x→0+))=xe−R/2. (E2) n(f ([n];x)) i=2 i The case x → 0− can be obtained from f([n];−x) = −f([n];x) and n(−x) = n(x). Inserting these expansions For the first term we will use the fact that h(x)δ(x(cid:48) −x) = inthedefinitionofthefunctionG(x)ofEq.(30)weseethat h(x(cid:48))δ(x−x(cid:48)), while for the second term we will use also itsderivativeisgivenby thatδ(g(x)−g(x(cid:48)))= 1 δ(x−x(cid:48)). Inparticularwehave g(cid:48)(x) x(cid:48) =f (f (x(cid:48))),andso,sincef(cid:48)([n],y)= n(y) : 2eR/2 i N−i+2 i n(fi([n];y) G(cid:48)(x→0+)= (cid:16) (cid:17) , (E3) x|R−ln 2 −ln(x)|3 δ(x(cid:48)−f ([n];x))=δ(f (f (x(cid:48)))−f (x)) 1+e−R i i N−i+2 i n(f ([n];x)) showing that G(x) has an infinite slope in x = 0. Further- =δ(f ([n];x(cid:48))−x) i . (D2) N−i+2 n(x) more,wealsohave,fromthepropertiesoftheco-motionfunc- tionandfromthesymmetryofthedensityn(−x)=n(x)that Plugging in this expression and using again the fact that h(x)δ(y−x)=h(y)δ(x−y)wefindthat G(−x)=2G(0)−G(x) (E4) G(f(x))=G(x)−G(0) forx>0 (E5) ∂2 G(f(x))=G(x)+G(0) forx<0. (E6) K ([n];x,x(cid:48))= ∂x∂x(cid:48) R Since f([n];−R/2) = R/2, for x = −R/2 both properties (cid:88)N w(cid:48)(cid:48)(|x−fi([n];x)|)δ(x(cid:48)−x) (E4)and(E6)musthold, implyingthatG(0) = 2G(−R/2), n(fi([n];x)) whichisEq.(37). i=2 When R is large, if x is well inside one of the atomic re- −(cid:88)N w(cid:48)(cid:48)(|x−x(cid:48)|)δ(x(cid:48)−fi([n];x)) gions then f∓([n];x) ≈ x ± R, yielding the constant dis- i=2 n(fi([n];x)) tance|x−fi([n];x)| ≈R,producingtheplateauxregionsin thekernel. Thisbehaviorholdsuntiltheelectroninf([n];x) =(cid:88)N w(cid:48)(cid:48)(|x(cid:48)−fi([n];x(cid:48))|)δ(x−x(cid:48)) approachestheoriginandstartsto“see”theseconddensityin n(f ([n];x(cid:48))) theoverlapregionpresentinthemidbond.Thishappenswhen i i=2 x≈±(R−1/a). Atthispoint,thelargenegativexbehavior −(cid:88)N w(cid:48)(cid:48)(|x−x(cid:48)|)δ(fN−i+2([n];x(cid:48))−x) ofG(x)startstoappear, n(x) i=2 1 G(x→−∞)= , (E7) =(cid:88)N w(cid:48)(cid:48)(|x(cid:48)−fi([n];x(cid:48))|)δ(x−x(cid:48)) n(0)x2 n(f ([n];x(cid:48))) i i=2 yieldingtheplateauvalueofEq.(34). −(cid:88)N w(cid:48)(cid:48)(|x−x(cid:48)|)δ(fN−i+2([n];x(cid:48))−x) n(f ([n];x(cid:48))) N−i+2 i=2 ∂2 = K ([n];x(cid:48),x), (D3) ∂x∂x(cid:48) R whereinthelaststepwejustrelabeledthesecondsum. 10 1 M.Seidl,Phys.Rev.A60,4387(1999). 22 N.T.Maitra,F.Zhang,R.J.Cave, andK.Burke,J.Chem.Phys. 2 M.Seidl,P.Gori-Giorgi, andA.Savin,Phys.Rev.A75,042511 120,5932(2004). (2007). 23 A. Dreuw and M. Head-Gordon, J. Am. Chem. Soc. 126, 4007 3 P.Gori-Giorgi,G.Vignale, andM.Seidl,J.Chem.TheoryCom- (2004). put.5,743(2009). 24 E.J.Baerends,O.V.Gritsenko, andR.vanMeer,Chem.Phys. 4 F. Malet, A. Mirtschink, J. C. Cremon, S. M. Reimann, and Phys.Chem.15,16408(2013). P.Gori-Giorgi,Phys.Rev.B87,115146(2013). 25 C. Bannwarth and S. Grimme, Comput. Theoret. Chem. 1040- 5 C. B. Mendl, F. Malet, and P. Gori-Giorgi, Phys. Rev. B 89, 1041,45(2014). 125106(2014). 26 M.DierksenandS.Grimme,J.Phys.Chem.A108,10225(2004). 6 F.Malet, A.Mirtschink, K.Giesbertz, L.Wagner, andP.Gori- 27 B. P. Peach, M. J. and, T. Helgaker, and D. J. Tozer, J. Chem. Giorgi,Phys.Chem.Chem.Phys.16,14551(2014). Phys.128,044118(2012). 7 F.Malet,A.Mirtschink,C.B.Mendl,J.Bjerlin,E.O.Karabulut, 28 T.Yanai,D.P.Tew, andN.C.Handy,Chem.Phys.Lett.393,51 S.M.Reimann, andP.Gori-Giorgi,Phys.Rev.Lett.115,033006 (2004). (2015). 29 W.J.Dreuw, A.andandM.Head-Gordon,J.Chem.Phys.119, 8 S.Vuckovic,L.O.Wagner,A.Mirtschink, andP.Gori-Giorgi,J. 2943(2003). Chem.Theory.Comput.11,3153(2015). 30 O.V.Gritsenko,S.vanGisbergen,A.Go¨rling, andE.J.Baerends, 9 A. Mirtschink, Energy Density Functionals from the Strong- J.Chem.Phys.113,8478(2000). Interaction Limit of Density Functional Theory, Phd thesis, VU 31 K.J.H.GiesbertzandE.J.Baerends,Chem.Phys.Lett.461,338 University,Amsterdam(2014). (2008). 10 A.Mirtschink,U.DeGiovannini,A.Rubio, andP.Gori-Giorgi, 32 J.F.Dobson,Phys.Rev.Lett.73,2244(1994). inpreparation(2015). 33 G.Vignale,Phys.Rev.Lett.74,3233(1995). 11 E.RungeandE.K.U.Gross,Phys.Rev.Lett.52,997(1984). 34 G.Vignale,Phys.Lett.A209,206(1995). 12 R.vanLeeuwen,Phys.Rev.Lett.82,3863(1999). 35 G.Buttazzo,L.DePascale, andP.Gori-Giorgi,Phys.Rev.A85, 13 M.RuggenthalerandR.vanLeeuwen,Eur.Phys.Lett.95,13001 062502(2012). (2011). 36 C.Cotar,G.Friesecke, andC.Klu¨ppelberg,Comm.PureAppl. 14 M.Ruggenthaler, M.Penz, andR.vanLeeuwen,J.Phys.Con- Math.66,548(2013). dens.Matter27,203202(2015). 37 M.Colombo,L.DePascale, andS.DiMarino,Can.J.Math.67, 15 J.I.Fuks,M.Farzanehpour,I.V.Tokatly,H.Appel,S.Kurth, and 350(2015). A.Rubio,Phys.Rev.A88,062512(2013). 38 M.ColomboandF.Stra,arXiv:1507.08522[math.AP]. 16 P.Elliott,J.I.Fuks,A.Rubio, andN.T.Maitra,Phys.Rev.Lett. 39 S.DiMarino,A.Gerolin,L.Nenna,M.Seidl, andP.Gori-Giorgi, 109,266404(2012). inpreparation. 17 J.I.FuksandN.T.Maitra,Phys.Chem.Chem.Phys.16,14504 40 F.MaletandP.Gori-Giorgi,Phys.Rev.Lett.109,246402(2012). (2014). 41 P.Hessler,J.Park, andK.Burke,Phys.Rev.Lett.82,378(1999). 18 S. Hirata and M. Head-Gordon, Chem. Phys. Lett. 302, 375 42 P.Hessler, N.P.Maitra, andK.Burke,J.Chem.Phys.117,72 (1999). (2002). 19 M.ThieleandS.Ku¨mmel,Phys.Rev.Lett.112,083001(2014). 43 Forframeswhicharetranslatedwithauniformvelocitywithre- 20 D.J.TozerandN.C.Handy,Phys.Chem.Chem.Phys.2,2117 specttothereferenceone,wesimplytalkaboutGalileaninvari- (2000). ance. 21 J.Neugebauer,E.J.Baerends, andM.Nooijen,J.Chem.Phys. 44 M. Mundt, S. Kummel, R. van Leeuwen, and P.-G. Reinhard, 121,6155(2004). Phys.Rev.A75,050501(R)(2007). 45 E.K.U.GrossandW.Kohn,Phys.Rev.Lett.55,2850(1985).

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.