Você está na página 1de 6

Vibrational partition functions for atomdiatom and atomtriatom van der Waals systems

Alice P. A. Urbano, Frederico V. Prudente, Antonio Riganelli and Antonio J. C. Varandas* Departamento de Quimica, Universidade de Coimbra, P-3049 Coimbra Codex, Portugal  Received 23rd May 2001, Accepted 12th September 2001 First published as an Advance Article on the web

The vibrational partition functions of atomdiatom and atom(linear triatom) van der Waals complexes are calculated directly within the classical statistical mechanics formalism by using a Monte Carlo procedure. Constrained classical Hamiltonians for these molecular systems have been derived by constraining the internal coordinates of the chemically bound molecules to their equilibrium values. The method is applied to the Ar CN and Ar HCN van der Waals molecules, and the results compared with previously reported ones.

1 Introduction
The most rigorous way to calculate the thermodynamical properties of a molecular system is via quantum statistical mechanics (QSM). This implies a detailed knowledge of its vibrational and rovibrational molecular spectra. Although rst-principle rovibrational calculations have seen a tremendous advance in recent years,14 the QSM approach still represents a formidable task which prevents its routine application. In recent work, we have shown57 that in some situations a viable alternative to QSM is classical statistical mechanics (CSM). In this approach, one evaluates the internal partition function by replacing the sum over states with the following multidimensional phase space integral8 Z Z 1 qint n expfbHq; pg dq dp 1 h b where H(q, p) is the classical Hamiltonian, h is the Planck constant, n is the number of degrees of freedom, (q, p) is the set of generalized coordinates and corresponding conjugate momenta, and b 1=kBT with kB being the Boltzmann constant. The subscript b in eqn. (1) implies that the hypervolume of integration is restricted to the phase space region associated to a bound situation: 0 H(q, p) De , where De is the classical dissociation energy of the molecule (see ref. 9). When used jointly with a Monte Carlo technique, such an approach yielded good results for the internal partition functions of both diatomic and triatomic van der Waals molecules.5,6 The theoretical investigation of both classical and quantum molecular properties starts with the formulation of the molecular Hamiltonian, which can be quite cumbersome for medium size and large molecules. However, many chemical processes can be described using simplied Hamiltonians which rely on a subset of the systems degrees of freedom. Such a subset consists of the so-called active coordinates, while the others (which are kept constant) are named inactive . The reduction of complexity is then achieved through holonomic constraints imposed on the Hamiltonian. Usually this procedure implies the algebraic inversion of large matrices which requires great eort. This diculty has been overcome by Hadder and Frederick10 who proposed a simple procedure to build holonomic constraints into the molecular Hamiltonian. Our aim in this work it to impose such holonomic constraints on the Hamiltonians of weakly bound complexes of the van der Waals type, and use them for calculations of the 5000 Phys. Chem. Chem. Phys., 2001, 3, 50005005

vibrational partition function. It should be noted that the vibrational frequencies associated with intramolecular bonds are typically one order of magnitude larger than those associated with intermolecular ones. The latter will then be taken as active degrees of freedom, while the intramolecular coordinates will be inactive . In other words, the system will be described by the coordinates associated with the weak vibrational frequencies. Examples of such molecules are the Ar CN and Ar HCN van der Waals (vdW) complexes which will be examined in the present work. These are wellknown prototypes of atomdiatom (X AB) and atomtriatom (X ABC) van der Waals systems, where the chemically stable species are the AB and linear-ABC molecules. To derive the classical Hamiltonian we will then employ the HadderFrederick procedure, while the vibrational partition functions will be calculated using eqn. (1). Valence coordinates (bond-lengths and included bond-angle) will be used to describe the vdW complexes. Although this choice may seem unadvantageous and somewhat arbitrary (particularly having in mind that the use of Jacobi coordinates may be self-suggesting), it has interesting consequences in the evaluation of the phase space integral, as we will discuss later. Note also that the two systems considered in this work may be considered as prototypes of more realistic situations involving larger polyatomics for which a formalism based on valence coordinates may be easier to rationalize the associated vibrational spectroscopy. In fact, valence coordinates are perhaps the most popularly used for such a purpose. Of course, for the species under consideration in this work, only two active coordinates will be considered: the distance between atom X and the diatomic or linear triatomic rigid systems, and the angle formed by this coordinate and the diatomic or linear triatomic axes. As a result, the integral in eqn. (1) has four dimensions. To solve it, we utilize a Monte Carlo technique initially suggested by Barker11 in the context of transition state theory, and subsequently employed by us for the calculation of molecular internal partition functions.57 The paper is organized as follows. Section 2 presents the general HadderFrederick method, which is there used to derive the classical Hamiltonians for the atom(rigid diatom) and atom(rigid linear triatom) vdW systems. In section 3, we summarize the details of the calculations and discuss the results obtained for the classical vibrational partition functions of Ar CN and Ar HCN. Some conclusions are in section 4. DOI: 10.1039/b104569p

This journal is # The Owner Societies 2001

The classical vibrational Hamiltonian

l Tvib Tint 1 T I 1 2

In a previous paper,6 we reported the classical vibrational Hamiltonian for atomdiatom systems using internal Jacobi coordinates. For the atomtriatom case, the amount of algebra involved in the procedure of Ref. 6 makes it extremely complicated. However, as pointed out in the Introduction, in simulations of large molecular systems it is common to use the concept of active and inactive (constrained) coordinates, i.e., only some coordinates (the active ones) contribute to the vibrational dynamics. This is the essence of adiabatic-type approaches. The procedure usually followed to obtain the classical Hamiltonian of a constrained system consists of the following steps (see Ref. 10 for further details): (a) Construction of the full classical vibrational Hamiltonian in terms of internal coordinates using the technique of Wilson and co-workers,1214 T h 1 pT G p 2 2

where and I are, respectively, the vibrational angular momentum vector and inertial tensor. Each term in eqn. (6) can easily be written in terms of the atomic Cartesian coordinates as N 1X _ _ Tint mi xi xi 7 2 i1
N X i1 N X i1

_ mi xi xi

Iab  Iab

mi xi xi dab ai bi

where p is the vector of conjugated momenta to the generalized internal coordinates vector (q), and G is the Wilson G matrix which has dimension (3N 6) (3N 6); N is the number of particles. (b) Transformation of the full vibrational Hamiltonian kinetic energy to its Lagrangian formulation, _ _ T l 1 qT M q 2 3

_ where q is the generalized velocity vector and M G1. (c) Imposition of the holonomic constraints in the full vibrational Lagrangian kinetic energy (by eliminating all rows and columns of the matrix M associated with these coordinates) to obtain the Lagrangian kinetic energy for the constrained system: _ _ T l 1 q0T M0 q0 2 4

where xi  (xi , yi , zi) is the position vector of atom i relative to the molecular axis frame, mi is the corresponding mass, and dab is the Kronecker delta symbol. The Lagrangian kinetic energy is obtained by dening the Cartesian coordinates in terms of the internal ones [i.e., xi xi(q)], followed by substitution of eqn. (7) to eqn. (9) into eqn. (6). The constraints are then imposed in the _ calculation of the velocities xi by requiring that the time derivatives of the constrained internal coordinates are zero. This leads to matrix M0 . It should be noted that the kinetic energy is a scalar quantity, and hence its value does not depend on the chosen molecular axis system. Thus, the most convenient frame of reference will be one that makes easiest the transformation between {xi} and q (a formal demonstration of this statement can be found in Appendix A of Ref. 10). The nal Hamiltonian kinetic energy term in eqn. (5) is obtained by inverting the matrix M0 . 2.1 Atom(rigid diatom)

The prime indicates that the system is now described using only active coordinates q0 . (d) Transformation of the constrained Lagrangian kinetic energy to its Hamiltonian formulation, T h 1 p0T G 0 p0 2 5

where G0 denotes the inverse of the reduced matrix M0 , and p0 is the vector of conjugated momenta. This procedure is dicult to carry out as it requires the algebraic inversion of the full matrix G. Hadder and Frederick10 proposed a simpler procedure to get directly the reduced matrix M0 and the Lagrangian kinetic energy term of the constrained system. Following their procedure, if L is the number of active coordinates, the eort to obtain the matrix G0 will be reduced to that of inverting a matrix of L L dimension. Not surprisingly therefore, the HadderFrederick method has been employed recently by several groups. For example, Goldeld15 used the HadderFrederick method to derive the quantum mechanical kinetic energy operator for cyclopropane in gas surface scattering studies, while Tretiakov and Cable16 employed it to obtain the classical kinetic energy term of diphenylamine. In turn, Vivian et al.17 used the Hadder Frederick procedure in a model study of torsionrotation interactions in toluene. In the following, we outline the key steps of the Hadder Frederick procedure to obtain the classical kinetic energy in terms of the internal active coordinates. We begin with the classical Lagrangian expression for the vibrational kinetic energy relative to an internal molecule-xed Cartesian coordinate system with the origin located at the center-of-mass of the system. It has the form

Fig. 1 shows the molecular axis frame adopted for the X AB system: r1 is the internuclear distance between atoms A and B, r2 the internuclear distance between atoms X and A, and y the angle between r1 and r2 . As the atoms move in a plane, we arbitrarily take this to be the xy plane with x being the axis parallel to the AB molecule and the origin located at the system center-of-mass. The rst step consists of determining the Cartesian components for the three atoms in terms of the coordinates (r1 , r2 , y). These are given by ! " mB # M r1 mX r2 cos y xA M mX r2 sin y yA M ! " mX mA # xB r1 mX r2 cos y M M 10 yB mX r2 sin y M ! " mB # m M r1 mBM A r2 cos y xX mB mA r2 sin y yX M

Fig. 1 System of valence coordinates used to describe a triatomic X AB van der Waals molecule; X is the rare gas atom.

Phys. Chem. Chem. Phys., 2001, 3, 50005005

5001

where M mA mB mX . In the next step, we obtain Tint , and I from eqn. (7), eqn. (8) and eqn. (9), respectively, while Tl is obtained from eqn. (6). Since the molecule is in the xy vib plane, only the z component of the internal angular momenta is nonzero. As a result, one requires to calculate only the zzcomponent of the inverse inertial momentum tensor, (I1)zz . Moreover, as Ixz Iyz 0, one gets (I1)zz 1=Izz . Taking the time derivatives of the Cartesian coordinates, where r1 is kept constant and r2 and y are considered to be active coordinates, one obtains after some algebra
2_ _2 2 Tint 1 m2 r2 1 m2 r2 y2 2 2 _ z m2 r2 m3 r1 r2 cos yy m3 r1 sin y_2 r 2 2 Izz m2 r2 m1 r1 2m3 r1 r2 cos y

11 12 13

planar. Thus, both the angle between r1 and r2 , and the torsional angle between the r1r2 and r2r3 planes need not be considered. Let us then assume xy to be the molecular plane, with the x axis being parallel to the ABC axis and the origin chosen to coincide with the system center-of-mass. Clearly, the procedure to obtain the kinetic energy term of the classical Hamiltonian for such a system is identical to that of the X AB case. From geometric considerations, one obtains for the Cartesian components of the four atoms in terms of the (r1 , r2 , r3 , y) coordinates: 3 ! 2 mB mC mX m xA r1 mCM X r2 mX r3 cos y M M 5 4 yA mX r3 sin y M xB yB xC yC xX yX ! ! 2 4
mA M r1

" mA
M

m r1 mCM X r2 mX r3 cos y M

where m1 (mB(mA mX))=M, m2 (mX(mA mB))=M, and m3 (mBmX)=M. Thus, Tvib in the Lagrangian formulation using internal coordinates assumes the form   2 1 m2 r1 sin2 y 2 l 3 _ m Tvib r2 2 2 Izz m r1 r2 sin y 3 m2 r2 m3 r1 cos y_2 y r _ Izz  r2 m2 r2 m2 r2 cos2 y 2 m2 2 2 3 1 Izz 2 Izz  2m m r1 r2 cos y _ 2 2 3 y 14 Izz _ _ Defining now q0 T (r_2 , y), the matrix M0 can be obtained by comparing eqn. (4) and eqn. (14). One then obtains G0 by inverting M0 . Finally, from eqn. (5) and using p0 T (Pr2 , Py), one obtains for the kinetic energy term in the Hamiltonian form
h Tvib

mX r3 sin y M
m mAM B r2 mX r3 cos y M

3 5 3 5

16

mX r3 sin y M
A

2m 4

m r1 mAM B r2 mA mB mC r3 cos y M mA mB mC r3 M

sin y

where M mA mB mC mX . Taking now the time derivatives of the Cartesian coordinates (r1 and r2 are kept constant, while r3 and y are considered to be active coordinates), one obtains after some algebra (we made extensive use of the Mathematica program18):
2_ _2 2 Tint 1 m4 r3 1 m4 r3 y2 2 2 _ z mX d sin y_3 m4 r3 mX dr3 cos yy r

17 18

1 m1 m2 m2 cos2 y 2 3 P 2 m2 m1 m2 m2 r2 3

m m2 r2 m3 r1 cos y sin y Pr2 Py 3 m2 r1 r2 m1 m2 m2 3

2 2 2 1 m2 m1 r1 m2 r2 2m3 r1 r2 cos y m2 r1 sin2 y 2 3 Py : 2 2 2 m2 m1 m2 m2 r1 r2 3

2 2 Izz m4 r3 mA r1 r2 2 mB r2 Md 2 2mX dr3 cos y 19

15 In the following section we employ this Hamiltonian [eqn. (15) plus the potential energy] to calculate the vibrational partition function of the Ar CN vdW complex. 2.2 Atom(linear rigid triatom)

where m4 (mX(mA mB mC))=M, and d (mAr1 (mA mB)r2)=M. Substituting these expressions in eqn. (6), one gets
l Tvib

  1 m2 d 2 sin2 y 2 mX dr3 sin ym4 r3 mX d cos y _ r3 m4 X 2 Izz Izz

For the case of X ABC, we dene the molecular axis frame as shown in Fig. 2: r1 is the internuclear distance between atoms A and B, r2 the internuclear distance between atoms B and C, r3 the internuclear distance between atoms C and X, and y the complement of the angle between r2 and r3 . As we assume the triatomic ABC to be linear, the complete system becomes

2 1 m r2 mX dr3 cos y m4 r3 2 _ 2 _ _ r3 y 4 3 20 y 2 Izz _ _ From eqn. (20) and defining q0 T (r_3 ,y), we can build as in the atomdiatom case the following M0 matrix: 0 M11 0 M21 0 M12 0 M22

M0 where

! 21

0 M11 m4 m2 X

d2 sin2 y Izz

22

0 0 M12 M21 mX

dr3 sin y m4 r3 mX d cos y Izz

23

  m2 r2 m mX dr3 cos y m2 d 2 0 2 M22 r3 m4 4 3 2 4 X cos2 y 24 Izz Izz Izz The final expression for the kinetic energy term in Hamiltonian form is then given by
Fig. 2 As is Fig. 1 but for a tetratomic X ABC van der Waals molecule.

 0 1 G11 h Tvib Pr3 ; Py G021 2

G012 G022



Pr3 Py

 25

5002

Phys. Chem. Chem. Phys., 2001, 3, 50005005

where G011 ,G012 ,G021 ,G022 are the elements of M0 1. For simplicity, we will not define explicitly the elements of matrix G0 because such expressions are complicated. However, they can be easily obtained using the following relations: M0 G011 0 0 22 0 0 26 M11 M22 M12 M21 G012 G021 M0 0 0 12 0 0 M11 M22 M12 M21 27

(5) Repeat steps 3 and 4 for i 3, . . . ,2n until the value (xS ) 2n of the last variable is sampled. This will result in a sampled point within the hypervolume of integration b dened in eqn. (1). (6) Calculate the weight factor for each sampled point g (xS ), according to g wg
2n Y i1

xmax xmin i i

32

G022

M0 0 0 11 0 0 M11 M22 M12 M21

28

which represents the hypervolume associated with xS . g (7) Repeat the steps 1 to 6 M times to calculate the integral in eqn. (1) as I % IM
M 1 X wg fg M g1

The classical Hamiltonian for the X ABC system will consist of the kinetic energy term given in eqn. (25) plus the potential energy. This Hamiltonian will be utilized in section 3 for the calculation of the vibrational partition function of Ar HCN.

33

where fg exp[ bH(xS )] is the function to be integrated. The g standard deviation associated with eqn. (33) is11 s2
M X 1 wg fg I M 2 MM 1 g1

Detailed calculations and results

34

We now present the vibrational partition function calculations for Ar CN and Ar HCN using the methods described above. Thus, we will evaluate the multidimensional phase space integral of eqn. (1) using the Monte Carlo method described elsewhere57 which employs a stratied and guided sampling procedure. The basic idea is to choose a sampling domain which resembles as much as possible the integration domain. According to this approach, the variables are not sampled independently. Instead, one establishes a hierarchy into the sampling procedure of each integration variable such as to optimize the eciency of the Monte Carlo technique. An important aspect to point out is that the Barker s algorithm, as described in the previous papers,6,7 assumes implicitly that there is no coupling between the conjugate momenta in the kinetic energy term. Thus, during the sampling procedure, the variables not yet sampled are kept xed in their equilibrium values. However, the kinetic energy employed in the present work has mixed terms and hence does not satisfy such a requirement, as seen from eqn. (15) and eqn. (26). As a result, a direct application of Barker s algorithm would lead to an integration domain smaller than the true one. It will then be necessary to generalize Barker s algorithm to overcome such a problem. Assuming x (q, p), the following procedure may be used: (1) Find the integration domain (xmin , xmax ) for the x1 vari1 1 able. xmin and xmax are, respectively, the smallest and largest 1 1 possible values that x1 can assume, irrespective of x2 ,. . .,x2n , according to: Hxmin ; x2 ; . . . ; x2n Hxmax ; x2 ; . . . ; x2n De 1 1 29

As will be shown in the present work, the order of the variables sampling is not crucial for systems with a single dissociation channel. We begin with the Ar CN vdW complex. The potential energy surface used to describe this system is a pairwise potential where the two-body terms have been represented by using the EHFACE19 model. Only one weak vdW minimum exists for such a species, with the calculated classical dissociation energy being De 0.000265858479 Eh . For further details, and numerical values of the parameters, the reader is addressed to refs. 20, 21. The vibrational partition function is calculated using two dierent expressions for the classical Hamiltonian. One is that of eqn. (15), which is written in terms of valence coordinates: the atoms Ar, C and N are labelled X, A, and B, respectively. The other is given in mass-scaled Jacobi coordinates (see, e.g., ref. 6) by   ! 1 1 1 HJ P2 2 2 P2 V R; r; f 35 f 2m R r R where R, r, and f are mass-scaled Jacobi coordinates (r is the scaled bound coordinate of diatom CN which is kept frozen at its equilibrium geometry, R the scaled distance between Ar and the center-of-mass of the diatom, and f is the Jacobi angle), and PR and Pf are the conjugate momenta of R and f; m ((mArmCmN)=M)1=2 is the reduced mass. In both cases, the CN fragment is kept constant at r 1.77237 a0 in eqn. (35), and r1 2.21490 a0 in eqn. (15), with 500 000 points being sampled to estimate the integral in eqn. (1). In this Monte Carlo calculation, R varies over the interval defined by Rmin 8.72015 a0 and Rmax 100.0 a0 for the Jacobi calculation, and rmin 7.05890 a0 and rmax 83.0 a0 for the calcula2 2 tion employing valence coordinates. Both Rmin and rmin are 2 the smallest possible values that R and r2 can assume for any values of f and y. To explain the importance of the hierarchy in the sampling scheme, we have calculated the vibrational partition function from eqn. (1) with the Hamiltonian of eqn. (15) by adopting two dierent strategies: one (qval ) consists of sampling rst the r2 r2 variable and then y, Pr2 and Py variables according to the algorithm reported elsewhere;7 the other (qval ) consists of y sampling rst the angle y and then r2 , Pr2 and Py . For the Jacobi J calculations (q ), we have used the hierarchy established in a previous paper,6 i.e. by sampling rst R, followed of f, PR and Pf . The results, together with their statistical uncertainties, are given in Table 1. Clearly, this shows that when using valence coordinates (qval and qval ) one obtains values in good agreement r2 y with those obtained by employing Jacobi coordinates, as it Phys. Chem. Chem. Phys., 2001, 3, 50005005 5003

(2) Sample randomly x1 within this range to obtain xS , 1 according to xS xmin xmax xmin x 1 1 1 1 30

where x is a random number between 0 and 1. (3) Starting with i 2, nd the integration domain (xmin , i max xi ) for the xi variable. xmin and xmax are, respectively, the 1 1 smallest and largest possible values that xi can assume with the rst (i 1) variables xed at the sampled values irrespective of xi 1 , . . . , x2n , according to: HxS ; . . . ; xS ; xmin ; xi1 ; . . . ; x2n 1 i1 i HxS ; . . . ; xS ; xmax ; xi1 ; . . . ; x2n De 1 i1 i 31

(4) Sample randomly the xi variable within this range to obtain xS as in step 2. i

Table 1 Vibrational partition functions for Ar CN using dierent classical Hamiltonians and sampling hierarchiesa T=K 10 20 30 40 50 60 70 80 90 100 200 300 400 500 600 700 800 900 1000
a

qJ 0.191 0.005 0.95 0.01 2.01 0.02 3.06 0.02 4.00 0.03 4.80 0.03 5.50 0.03 6.09 0.04 6.60 0.04 7.04 0.04 9.48 0.05 10.49 0.06 11.04 0.06 11.38 0.06 11.62 0.06 11.79 0.06 11.92 0.06 12.03 0.07 12.11 0.07

qval r2 0.188 0.005 0.94 0.01 2.00 0.02 3.06 0.02 4.00 0.03 4.81 0.03 5.50 0.04 6.10 0.04 6.61 0.04 7.05 0.04 9.50 0.05 10.51 0.06 11.06 0.06 11.40 0.06 11.64 0.06 11.81 0.07 11.94 0.07 12.05 0.07 12.13 0.07

qval y 0.195 0.006 0.94 0.01 1.98 0.02 3.02 0.02 3.95 0.03 4.75 0.03 5.43 0.04 6.02 0.04 6.53 0.04 6.97 0.05 9.39 0.06 10.40 0.06 10.95 0.06 11.29 0.07 11.52 0.07 11.70 0.07 11.83 0.07 11.93 0.07 12.01 0.07

In this and in Table 2, the statistical errors are included.

should. A comparison between qval and qval also shows that the r2 y sampling order is irrelevant in the present case. The various J results (q , qval , and qval ) nearly coincide within their statistical r2 y uncertainties. It is important to point out that there is a small dierence between the qJ results presented here and those reported in previous work (ref. 6). This stems from a dierence in the accuracy of the dissociation energy employed for the two calculations: previously6 we have used De 0.000266 Eh which diers slightly from the value quoted above. We now turn to the vibrational partition function of Ar HCN using the classical Hamiltonian derived in section 2.2; X, A, B, and C denote the atoms Ar, H, C, and N, respectively. The potential energy surface used to describe this system is the energy switching22 (ES) function reported by Varandas et al.23 (denoted ESIII in their paper). This ESIII potential energy surface was obtained by merging a DMBE potential energy surface20 with a Legendre analysis t to ab initio energies calculated by Clary et al.24. As reported in ref. 23, the ESIII potential energy surface has one global and three local minima, with the classical dissociation energy being accurately given by De 0.00038247951 Eh . In this case, the
Table 2 Vibrational partition functions for Ar HCN using dierent sampling hierarchies T=K 10 20 30 40 50 60 70 80 90 100 200 300 400 500 600 700 800 900 1000 qval r2 0.099 0.002 0.621 0.005 1.570 0.008 2.72 0.01 3.90 0.01 5.02 0.02 6.05 0.02 6.97 0.02 7.81 0.02 8.55 0.02 13.06 0.03 15.10 0.04 16.24 0.04 16.98 0.04 17.49 0.04 17.86 0.04 18.15 0.05 18.38 0.05 18.56 0.05 qval y 0.094 0.002 0.610 0.005 1.554 0.009 2.70 0.01 3.88 0.02 4.99 0.02 6.01 0.02 6.94 0.02 7.76 0.03 8.51 0.03 12.99 0.04 15.03 0.04 16.17 0.04 16.90 0.05 17.41 0.05 17.79 0.05 18.07 0.05 18.30 0.05 18.48 0.05

results obtained for the vibrational partition function calculated using the Monte Carlo procedure could be more sensitive to the hierarchy used for sampling the variables. For this reason, we have calculated the vibrational partition function from eqn. (1) (with the classical Hamiltonian expressed in terms of valence coordinates) by adopting two dierent schemes. One (qval ) consists of sampling rst the r3 variable, followed by y, Pr3 r3 and Py . In the other (qval ), one samples rst y, followed by r3 , y Pr3 , and Py . Both calculations have been carried out by keeping the HCN triatomic frozen at its equilibrium linear geometry (r1 2.01600081 a0 and r2 2.19200 a0). A total of 3 106 phase space points has been sampled to evaluate the integral in eqn. (1). Moreover, we have set rmin 6.2920253 a0 , and 3 rmax 100.0 a0 in order to cover fully the bound phase space. 3 The results, and their associated statistical errors, are reported in Table 2. It is seen that the order in which the variables are sampled also has in this case only a negligible inuence. In fact, the qval and qval results coincide within their statistical errors. r3 y The only detail which we must bear in mind concerns the value of rmin used for the qval calculation: it must be the smallest 3 r3 optimum value of r3 , irrespective of y. Finally, we assess the accuracy of the present results by using a denition of reduced temperature proposed by us.6 Based on such a denition, it has been proposed that classical statistical mechanics can yield accurate results of the internal partition function of polyatomic systems described by realistic potentials for temperatures above T e0=kB , where e0 is the zero point energy of the system. Specically, the deviation of the classical partition function from the quantum one has been estimated to be smaller than 10% above that temperature. Thus, considering that e0 16.859 cm1 for Ar CN and e0 26.526 cm1 for Ar HCN, the results presented in Tables 1 and 2 should provide reliable estimates of the quantum vibrational partition function for the Ar CN and Ar HCN vdW systems at temperatures above T 24.26 K and T 38.17 K, respectively.

Conclusions

We have derived the classical vibrational Hamiltonian for two constrained systems [atom(rigid diatom), and atom(rigid linear triatom)] by using the general procedure proposed by Hadder and Frederick.10 These should be of help for several applications. As an example, we have utilized such Hamiltonians for the calculation of the vibrational partition functions of Ar CN and Ar HCN vdW complexes. For this, we have employed an ecient Monte Carlo approach to solve the multidimensional phase space integrals which arise in the classical statistical mechanics formulation. The values obtained for Ar CN are close to those reported using the classical Hamiltonian expressed in Jacobi coordinates. We have also veried that the order through which the variables are sampled is not crucial provided that the vdW system has only one dissociative channel.

Acknowledgements
This work has the support of Fundacao para a Ci^ncia e e Tecnologia, Portugal, under programme PRAXIS XXI. FVP also acknowledges partial nancial help through Fundacao Coordenacao de Aperfeicoamento de Pessoal de Nivel Super  ior (CAPES, Brazil).

References
1 J. Tennyson, Comp. Phys. Rep., 1986, 4, 1. 2 Z. Baci and J. C. Light, Annu. Rev. Phys. Chem., 1989, 40, 469. c

5004

Phys. Chem. Chem. Phys., 2001, 3, 50005005

3 T. Carrington, Jr., in Encyclopaedia of Computational Chemistry, John Wiley & Sons, New York, 1998, p. 3157. 4 F. V. Prudente, L. S. Costa and P. H. Acioli, J. Phys. B: At. Mol. Opt. Phys., 2000, 33, R285. 5 A. Riganelli, W. Wang and A. J. C. Varandas, J. Phys. Chem. A, 1999, 103, 8303. 6 A. Riganelli, F. V. Prudente and A. J. C. Varandas, Phys. Chem. Chem. Phys., 2000, 2, 4121. 7 F. V. Prudente, A. Riganelli and A. J. C. Varandas, J. Phys. Chem. A, 2001, 105, 5272. 8 L. Landau and E. Lifshitz, Statistical Physics, Pergamon Press, New York, 1969. 9 A. Riganelli, F. V. Prudente and A. J. C. Varandas, J. Phys. Chem. A, 2001, 105, 9518. 10 J. E. Hadder and J. H. Frederick, J. Chem. Phys., 1992, 97, 3500. 11 J. R. Barker, J. Phys. Chem., 1987, 91, 3849. 12 E. B. Wilson, Jr. and J. B. Howard, J. Chem. Phys., 1936, 4, 260. 13 E. B. Wilson, Jr. and B. L. Crawford, J. Chem. Phys., 1938, 6, 223.

14 E. B. Wilson, Jr., J. Chem. Phys., 1939, 7, 1047. 15 E. M. Goldeld, Faraday Discuss., 1998, 110, 185. 16 I. V. Tretiakov and J. R. Cable, J. Chem. Phys., 1997, 107, 9715. 17 J. T. Vivian, S. A. Lehn and J. H. Frederick, J. Chem. Phys., 1997, 107, 6646. 18 http:==www.wolfram.com. 19 A. J. C. Varandas and J. D. Silva, J. Chem. Soc., Faraday Trans., 1992, 88, 941. 20 S. P. J. Rodrigues and A. J. C. Varandas, J. Phys. Chem. A, 1998, 102, 6266. 21 A. J. C. Varandas and S. P. J. Rodrigues, J. Chem. Phys., 1997, 106, 9647. 22 A. J. C. Varandas, J. Chem. Phys., 1996, 105, 3524. 23 A. J. C. Varandas, S. P. J. Rodrigues and P. A. J. Gomes, Chem. Phys. Lett., 1998, 297, 458. 24 D. C. Clary, C. B. Dateo and T. Stoecklin, J. Chem. Phys., 1990, 93, 7666.

Phys. Chem. Chem. Phys., 2001, 3, 50005005

5005

Você também pode gostar