some comments on quasi-bound-state calculations

3
Volume 65A, number 3 PHYSICS LETTERS 6 March 1978 SOME COMMENTS ON QUASI-BOUND-STATE CALCULATIONS J. KILLINGBECK Physics Department, University of Hull, UK Received 13 December 1977 Quasi-bound-state energy calculations are made for a perturbed oscillator and a perturbed hydrogen atom. The theory of the least-squares method is analysed. A simple numerical variant of the well-known stabilization method is tested, and the results are compared with those from the least-squares and perturbation-series methods. When an unbounded perturbing operator (e.g. the Various arguments have been presented to justify Stark-effect operator Xz) is added to a hamiltonian the least-squares approach. but it seems to us that their which has bound states, the bound states are all de- essence can be summarized as follows. The work of stroyed in principle. For small X a time-dependent ap- Titchmarsh [9] and others has shown that adding a proach suggests that a particle initially in some bound term such as Xz or —-~XIr to the hydrogen-atom harnil- state will only tunnel through slowly to the “outer” tonian gives a hamiltonian H for which the Green’s region of the potential. Several time-independent ap- function G(x, x’, E), the kernel of the integral opera- proaches to this kind of problem have been suggested. tor representation of the resolvent operator (H —- E~ I Reinhardt [1] made a numerically successful applica- has poles just below the real L’ axis. To look for these tion to the Stark effect of a technique which had only poles we can evaluate R(~, F) = ~~(H—EY 210>,vary- been rigorously justified for dilation analytic poten- ing ~ and F, to maximize R(~, F). The optimum Ewill tials [2], and which involves finding complex eigen- be the real part of the complex-pole energy, since R values for a “rotated” hamiltonian. Methods not in- will have a maximum as we pass by the pole along the volving complex eigenvalues include the least-squares real axis. From the Schwartz inequality, or from a approach [3,41, the stabilization method using secular more elaborate variational argument. we find that determinants [5,6], and the method of ref. [7], which ~(H—E)2fr/)~ is a lower bound on R(ç5, F). Also, uses an extra stabilizing potential to convert the quasi- for a given ~, the optimum F is clearly (~IHIØ~, so we bound states into true bound states. The above rneth- are led to minimize ods lead to calculations quite similar to those for tra- ditional bound states; in the present note we report ~2(~) = (~H2I~) (~H~)2. (1) some formal and numerical work relating to the least- by varying ci, taking (c5~HI~> as the optimum F value. squares and stabilization methods. If the harniltonian H can be given a ci-dependent par- In ref. [3] results are quoted for hydrogen-atom titioning H = H 0 + V. so that ~ is an eigenfunction of Stark-effect eigenvalues obtained by the Rayleigh— H0, then ~ 2(ci) takes the simple form (~ V~I~) Ritz method. We note here that, even with the two simple basis functions e~” and ze°”, varying ~ For the perturbed oscillator hamiltonian of Bender leads to an eigenvalue tending to minus infinity for and Wu [10] the partitioning gives the following re- any non-zero value of X [8]. To produce finite eigen- sults, with ~ = e_~x2 (normalized), values a must be constrained (at least in the e c.!r function) and the choice a = 1 is suggested by pertur- H= (—D2 + 4a2x2) + (~Xx4 + [~ 4a2]x2), (2) bation theory. 180

Upload: j-killingbeck

Post on 21-Jun-2016

215 views

Category:

Documents


3 download

TRANSCRIPT

Volume 65A, number3 PHYSICSLETTERS 6 March1978

SOME COMMENTS ON QUASI-BOUND-STATE CALCULATIONS

J. KILLINGBECKPhysicsDepartment,Universityof Hull, UK

Received13 December1977

Quasi-bound-stateenergycalculationsaremadefor a perturbedoscillator anda perturbedhydrogenatom. Thetheory oftheleast-squaresmethodis analysed.A simplenumericalvariantof the well-known stabilizationmethodis tested,andtheresultsarecomparedwith thosefrom the least-squaresand perturbation-seriesmethods.

When an unboundedperturbingoperator(e.g.the Variousargumentshavebeenpresentedto justifyStark-effectoperatorXz) is addedto a hamiltonian theleast-squaresapproach.but it seemsto us that theirwhichhasboundstates,theboundstatesareall de- essencecan be summarizedasfollows. The work ofstroyedin principle. For small X a time-dependentap- Titchmarsh[9] andothershasshownthat addingaproachsuggeststhat aparticle initially in somebound termsuch asXz or —-~XIrto thehydrogen-atomharnil-statewill only tunnel throughslowly to the “outer” tonian givesa hamiltonianH for which the Green’sregionof thepotential. Severaltime-independentap- function G(x, x’, E), the kernel of theintegralopera-proachesto this kind of problemhavebeensuggested. tor representationof theresolventoperator(H —- E~I

Reinhardt [1] made anumericallysuccessfulapplica- haspolesjust belowthe real L’ axis. To look for thesetion to the Starkeffectof a techniquewhichhadonly poleswe can evaluateR(~,F) = ~~(H—EY210>,vary-beenrigorouslyjustified for dilation analytic poten- ing ~ andF, to maximizeR(~,F). The optimumEwilltials [2], andwhichinvolves finding complexeigen- be the real part of thecomplex-poleenergy,sinceRvaluesfor a “rotated” hamiltonian.Methodsnot in- will havea maximumaswe passby thepole alongthevolving complexeigenvaluesinclude theleast-squares real axis. From theSchwartzinequality, or from aapproach[3,41,thestabilizationmethod usingsecular more elaboratevariationalargument.we find thatdeterminants[5,6], andthemethod of ref. [7], which ~(H—E)2fr/)~ is a lower boundon R(ç5, F). Also,usesan extra stabilizingpotential to convert the quasi- for agiven ~, theoptimumF is clearly (~IHIØ~,so weboundstatesinto true boundstates.The aboverneth- areledto minimizeods leadto calculationsquite similar to thosefor tra-ditional boundstates;in thepresentnote we report ~2(~) = (~H2I~) (~H~)2. (1)someformal andnumericalwork relatingto the least- by varyingci, taking (c5~HI~>astheoptimumF value.

squaresandstabilizationmethods. If theharniltonianH canbe givenaci-dependentpar-In ref. [3] resultsarequoted for hydrogen-atom titioning H = H

0 + V. so that ~ is an eigenfunctionofStark-effecteigenvaluesobtainedby the Rayleigh— H0, then ~

2(ci) takesthesimpleform (~V~I~)Ritz method.We note herethat,evenwith the twosimplebasisfunctionse~”andze°”, varying~ For theperturbedoscillatorhamiltonianof Benderleadsto an eigenvaluetending to minus infinity for andWu [10] thepartitioning givesthefollowing re-

any non-zerovalueof X [8]. To producefinite eigen- sults,with ~ = e_~x2(normalized),valuesamustbe constrained(at leastin the e c.!r

function) andthechoicea= 1 is suggestedby pertur- H= (—D2+ 4a2x2)+ (~Xx4+ [~— 4a2]x2), (2)bation theory.

180

Volume 65A, number3 PHYSICS LETTERS 6 March1978

Table 1Perturbed-oscillatorresults.

x 0.05 —0.01 —0.03 —0.05

E(leastsquares) 0.5331 0.4923 0.4750 0.4548E (numerical) 0.5327 0.4922 0.4742 0.4507E (perturbation) 0.5332 0.4922 0.4742 0.4507IdE/dXI (minimum) — (<2 X 10—8) 1.0 X 10~ 5.0 X i0~

Table 2Perturbed-hydrogen-atomresults.

x 0.05 —0.02 —0.04 —0.06

E (least squares) —0.4266 —0.5303 —0.5610 —0.5923E (numerical) —0.4282 —0.5307 —0.5631 —0.5983E(perturbation)a) —0.4281 —0.5307 —0.5630 —0.5979IdE/dXI (minimum) — (<5 X 10~) 4.8X 10_6 40x i0~

a) Fifth order sums.

3 A2 1 1 3 x Table3= + — (-4 — 4a2)2 + ~— (4 —4a2), Nodepositionsfor oscillator (X = —0.01). (Integrationstrip

~ a~ ° a2 ~ a3 widthh = 0.02.)(3)

E Nodepositions

(~IHIci)= ~ ~ (64a3 + 4a+ 3X). (4)64 a2 0.4924 4.57, 11.21, 12.090.4923 4.75, 11.21, 12.09

The correspondingresultsfor the perturbedhydrogen- 0.4922 — 11.21, 12.09

atomproblemof Titchmarsh[9], with 0 = e~-”(nor- 0.4921 — 11.21 12.09

malized)are:

H = (_~D2— 1 D — ~‘~+ (xr + ~ ~ (5) Tables1 and2 also includenumericalresultsob-r ri \ r f tamedasfollows. A realE wasguessed,andthe Schrö-

2 dingerequationwas integratedoutfrom the origin by= a2(l — a)2 + X(l —a)+ ~h—, (6) the methodof ref. [11], the nodepositionsbeing

~a2 noted.VaryingEgives resultssuchasthoseof table3,3 ~ in which the first nodeis seento movevery rapidly

(OIHIO)= ~a2 — a +~—. (7) from 4.75 to 11.21 asE changesby 0.0001.Thisshowsthat theenergyis changingvery slowlywith X,

The resultsobtainedby minimizing ~2 are shown in if we solvetheSchrodingerequationfor theDirichlettables 1 and 2. A positive A valueis includedasa condition~ = 0 at x = X. In their discussionof theroughcheckon thetrial functions.For A> 0, (OIHIO) stabilizationmethod,HaziandTaylor [6] arguethat,— ~)givestheTemplelowerboundon the bound- by adding functionsto a basisset until the eigenvaluestateenergy.(For the Starkeffecta few preliminary stabilizes,theyare effectivelydoing the Dirichiet prob-calculationssuggestthat the Green’sfunction pole lem for graduallyincreasingX values.Wehaveherelies within a semi-circleof radius~(Ø)about(OIHIO), solved theDirichiet problemdirectly, which is farbut it is not clearwhetherthis exemplifiesa general more easythan doingthematrix calculation! Thetheorem.) quantitydE/dX is well defined,whereasthe quantity

181

Volume 65A, number3 PHYSICS LETTERS 6 March 1978

dE/dN(whereN is thenumberof basisfunctions)is mit theF dependenceof phaseshiftsto be evaluated.not; Hazi andTaylor’s discussionsuggeststhat the Why thevariousmethodsagree,andwhat theygiveminimum value of sucha derivateshould be roughly us (we hopeit to be thereal part of theGreen-func-proportionalto theimaginarypartof thepole posi- tion pole position) is, despitemuch discussion[3,6,9],tion. not completelyclear.We hopethat theideasandre-

If the traditionalRS perturbedfunctions (cio, ~ suits outlinedherewill stimulatefurther discussion.

areusedas thebasisset,we knowthat thepower-seriesexpansion(in A) of the eigenvaluecorre- Referencesspondingto cio must agreeto (2n + l)th orderwiththe RS energyseries.Accordinglywe might expect Ill W.P. Reinhardt,Intern.J. QuantumChem.Sym. 10

theenergyseriesto exhibit a stabilizationeffect. This (1976)359.is just what variousasymptoticseriesdo! Tables1 and 121 B. Simon,Ann. Math. 97(1973)247.2 showtheresults(to four figures)of summingthe 131 P. FroehichandE. Brandas,Phys.Rev. A12 (1975)1.141 F.H. Read,Chem.Phys.Lett. 12 (1972)549.RS seriesup to its smallestterm. (Incidentally,the [51I. Ehiezer,H.S. TaylorandJ.K. Williams, J.Chem.Phys.hydrogen-atomStark-effectseriesbeyondA4 doesnot 47 (1967)2165.seemto havebeenevaluated).The agreementbetween 161 A.U. Haziand H-S. Taylor, Phys.Rev. Al (1970)1109.the resultsof thevariousmethodsis remarkable,if we 171 J.F. Liebman,D.Y. YeagerandJ.Simons,Chem.Phys.

Lett. 48 (1977)227.rememberthe simple form of the least-squarestrial 181 J.Killingbeck, Rept. Progr.Phys.40 (1977)963.function.We also found closeagreementbetweenthe [91 E.C. Titchmarsh,Proc. Roy. Soc.A210 (1952)30.numericalandleast-squareresultsfor thepotential [101C. BenderandT.T. Wu,Phys. Rev. 184 (1969) 1231.

2x2e Kx2, wherethe oscillationsat largex also per- Ill] J. Kihlingbeck,J.Phys.AlO (1977)L99.

182