8/3/2019 Oleg L. Polyansky and Jonathan Tennyson- On the Convergence of Effective Hamiltonian Expansions

1/6

JOURNAL OF MOLECULAR SPECTRO SCOPY 154,246-25 l ( 1992)

On the Convergence of Effective Hamiltonian ExpansionsOLEG L. POLYANSKY

Institute of Applied Physics, Academy of Sciences of the USSR, 46 U(ianov Street,603600 N izhn y Novgorod, Rus sia

ANDJONATHAN TENNYSON

Department of Phy sics and Ast ronomy, Univ ersity College London,Gow er Street, London W ClE 6BT, U nited Kingdom

The superiorityof the convergencebehaviorof effective otationalHamiltonians xpressed interms of Pade approximants over more conventional power series Hamiltonians is well known.For diatomics it is suggested that the failure of the conventional expansions is due to singularitiesin the complex Jplane. This theory is supported by numerical calculations on rotationally excitedHa with both real and imaginary J. These calculations show that the Pade expansions representthe data 30 times better than a power series and locate a singularity at J - 25 for negative J(J+ 1) The rotational structure of quasilinear molecules is found to have a more serious singularitycorresponding to that of the Bore1 Hamiltonian. The implications of these findings for effectiveHamiltonian expansions are discussed. 0 1992 Academic PIW, Inc .

1. INTRODUCTIONEffective Hamiltonians are very widely used as a means of representing molecular

rovibrational spectroscopic data. For rotational levels, these effective Hamiltoniansare conventionally expressed as power series expansions in terms of the quantumnumbers J( J + 1) and K2 ( I ) . For light molecules and molecules with large amplitudevibrational motion such power series are often poorly convergent. In some cases, suchas water (2, 3), the power series is known to actually diverge.

As a solution to this problem, other effective Hamiltonian expansions have beensuggested. Suggestions, based on the study of quartic perturbed harmonic oscillators(4, 5), include the use of Pad& and Bore1 approximants (3). These have been foundto give greatly improved representations of the spectroscopic data of water (3, 6),H2D+ ( 7,8), and C203 (9) when compared to power series expansions with the samenumber of parameters.

The purpose of this paper is to give a theoretical justification, if not a formal proof,of the superiority of the Pad& and Bore1 effective Hamiltonian representations of theexact energy eigenvalues of molecular rovibrational Hamiltonians. This will be doneby considering the behavior of the power series expansions in the complex (J, K)plane and in particular by considering the cases J( J + 1) < 0 and K2 < 0.

Formally, the convergence properties of a given series expansionf(x) = g a,x (1)

n=O

0022-2852192 $5.00Copyright 0 1992 by Academic Press, Inc.All rights of reproduction in any form reserved.

246

8/3/2019 Oleg L. Polyansky and Jonathan Tennyson- On the Convergence of Effective Hamiltonian Expansions

2/6

CONVERGENCE OF EFFECTIVE HAMILTONIANS 247

are determined by the radius of convergence, p, of the series. The series is convergentfor all 1x 1 < p and divergent for all 1XI > p. For 1xl = p the series may eitherconverge or diverge. Furthermore, p is simply equal to the distance to the nearestsingularity in the function f( x) in the complex plane.A simple example is provided by the series+x= (-1)x.n=OAs this function is singular at x = - 1, p = 1 for this series.

In this work we consider two representative one dimensional problems: (a) therotational levels of a diatomic molecule, for which the usual series expansion is interms of J(J + 1), and (b) the rotational levels of a quasilinear molecule with abending coordinate, for which a power series in K* is often used.

2. DIATOMIC MOLECULESFor a diatomic molecule the Schrodinger equation is given by

2+;:+2p E- V(r)- 2pr2[J(J+ 1)

I* = 0,and solutions are characterised in terms of the rotational quantum number J. Per-turbative solutions of Eq. ( 3) give a power series in J( J + 1),E(J) = BJ(J+ 1) - DJ*(J+ l)* + HJ3(J+ 1)3

- LJ4(J+ 1)4 + PJ(J+ 1)5, (4)were terms up to 10th order have been retained.Table I gives values for these constants for H2 determined by fitting the observed

(10) rotational levels of the vibrational ground state and pure rotational data (II,12). It is notable that the ratios between neighboring constants in the power seriesexpansion are all similar (to within 30%)

TABLE IRotational Constants (in cm-) for HZ

I II III IV

B 59.330075(21) 59.334604(22) 59.334633(M) 59.334455(l)10 D 4.54710(3) 4.56890(l) 4.56918(11) 4.566436(3)10H 4.4999(3) 4.88853(2) 4.8997(29) 4.89390(3)l@ L 3.858(7) 6.139(2) 6.343(32) 6.5088(7)lOlOP 0.155(4) 0.546(S) 0.716(12) 0X12(4)

(5)

Note. (I) Power series fit of al1 aboratory data of Refs. (IO-12), (II) power series fit to the rotational dataup to J = 15, (III) fit of Jennings et al. to their rotational and corrected Orion data using a power serieswith geometric corrections (12), (IV) Pad6 fit to all the laboratory data. (Values in parentheses give onestandard deviation in units of the final digit).

8/3/2019 Oleg L. Polyansky and Jonathan Tennyson- On the Convergence of Effective Hamiltonian Expansions

3/6

248 POLYANSKY AND TENNYSONThis behavior is suggestive of a geometric series for which the ratio of neighboringterms is a constant.

The equivalent Pad6 representation to the power series (4) is given byE(J) = (BH* - BDL)J(J+ 1) - aJ(J+ l)* + pJ3(J+ 1)3H* - DL + (LH - DP)J(Jf 1) + (L* - HP)J*(J+ l)*

CY= -BHL -I- DBP - D2L + DH2,@ = 3L2 - 2DHL -I- D=P + H3 - BHP. (6)

Table I presents four different fits to the available rotational levels of the vibrationalground state of HZ. The columns in Table I are given in order of increasing accuracyof the fits. The fits to all data (columns I and IV) were obtained by weighting therotational data of Refs. ( 11, 12) 1000 times higher than the rovibrational data of Ref.( IO). The Pad& formula (column IV) reproduces these data with a standard deviation30 times smaller than the polynomial expansion of Eq. (4) (column I).

Indeed the observed H2 data are reproduced more accurately than the power seriesexpansion by the simpler, sixth-order Pad& expression

E(J)=BJ(J+ l)- DJ2(J+ I)*1 + (H/D)J(J+ 1) (7)If J( J + 1) is replaced by - J( J + 1) in Eq. (7), then this formula passes through asingularity near J = 25 for the rotational constants of Table I. Furthermore, thegeometric properties of Eq. (5) means that this singularity will persist at higher, possiblyall, orders.Until now, the use of Pad& rather than polynomial expansions has been justifiedon purely phenomonological grounds. However the observations above suggest thatthe exact solutions of Eq. (3) have a singularity for J - 25 if one takes J( J + 1) z , and HNCS have often been observed to have a factorial,n!, behavior. This behavior is clearly mimicked by the Bore1 representation.

4. CONCLUSIONSWe have suggested that the convergence properties of effective Hamiltonians need

to be analyzed in terms of their behavior in the complex plane. Unlike conventionalpolynomial expansions, the behavior of two effective Hamiltonians, the Pad& withJ( J + 1) and the Bore1 with K*, is shown to be in accord with the properties of therotational levels they are designed to represent. This gives substantiation for the sug-gestion (3) that these Hamiltonians be used and an explanation for the superior fitsgenerally obtained with these Hamiltonians as compared to more conventional models.

Perhaps the most challenging case to analysis is for asymmetric tops with largeamplitude motion. We believe that analysis similar to that applied here can be used

8/3/2019 Oleg L. Polyansky and Jonathan Tennyson- On the Convergence of Effective Hamiltonian Expansions

6/6

CONVERGENCE OF EFFECTIVE HAMILTONIANS 251to explain the anomalous behavior of the J( J + 1) power series for molecules suchas Ar-HCN ( 17) and N2F4. We are presently studying the Ar-HCN case for whichthe critical value of J appears to be as low as 10. The study of the singularities in suchmolecules will lead us to better effective Hamiltonians, better fits of the observed data,and hence a better understanding of the behavior of molecular energy levels as afunction of rotational quantum numbers.

ACKNOWLEDGMENTWe thank The Royal Society for a grant which allowed O.L.P. to visit London during the course of

this work.

RECEIVED: December 26, 1991REFERENCES

1. M. R. ALIEV AND J . K. G. WATSON , in Molecular Spectroscopy: Modem Research (K. NarahariRao, Ed.), Vol. III, p. 1, Academic Press, Orlando, 1985.

2. C. CAMY-PE YRET AND J.-M. FLAUD, Mol . P hys . 32,499-521 ( 1976).3. 0. L. P OLYANSKY, J. Mol. Spectrosc. 112, 79-87 ( 1985).4. B. SIMON,Ann. Phys. 58,76-136 ( 1970).5. S. GRAF FI ,V. GREC HI, AND B. SIMON, Whys . Left. B 32,631-634 ( 1970).6. S. P. BEL OV, I. N. KOZIN, 0. L. POL YANSKY, M. Yu. TR ETYAKOV, AND N. F . Zom v, J . Mol. Spectrosc.

126, 113-117(1987).7. S. C. FOS TER , A. R. W. MCKE LLAR, 1. R. P ETE RKI N, . K. G. WATSON, F. S. P AN, M. W. CR OF TON,

R. S. ALTMAN, AND T. OKA, J. Chem. Phy s. 84,91-99 (1986).8. I. N. KOZIN, 0. L. P OLYANSKY, AND N. F . ZOBOV, J. Mol. Spectrosc. 128, 126-l 34 ( 1988).9. J. VANDER ALJ WERA,0. L. P OLYANSKY, AND J . W. C . J OHN S, J. Chem. Phys. 95,2299-23 16 ( 1991).10. I . DABROWSKI, Can . J. Phys. 62, 1839-1664 (1984).

11. D. E. J E N N I NG S ,S. L. BR AGG, AND J . W. BR AULT, As troph ys . J. 282, L85-L88 ( 1984).12. D. E. J E N N I NG S ,S. A. R AHN, AND A. OWYOUN G, Astrophys. J. 291, L15-L18 (1985).13. R. J. LE ROY, J . Chem. Phys. 54,5433-5434 ( 1971).14. D. M. BISHOP AND S.-K. SHI H, J. Chem. Phys. 64, 162-169 (1976).15. W. R. THO RS ON AND I. NAKAGAWA, J. Chem. Phys. 33,994-1004 ( 1960).16. S. R. POLO, J. Mol. Spectrosc. 23, 117-l 18 (1967).17. R. E. BUMGARNER AND G. E. BLAKE, Chem. Phy s. Left . 161,308-314 ( 1989).