Você está na página 1de 8
Z Phys. D 35, 107-118 (1995) ZEITSCHRIFT FUR PHYSIKD © Springer-Verlag 1995 Classical theory of vibration energy transfer in collinear collisions H. Skenderovi!, $. Danko Bosanac” “nsttute for Physics, Bijenika c, 50, $1001 Zagreb, Republic of Croatia PR. Boskovié Institute, Bijniéka 54, $1001 Zagreb, Republic of Croatia Received: 2 January 1995/ Final version: 4 April 1995 Abstract. Classical theory of collisions is east in a form which also includes the uncertainty principle, This theory is used for analyzing the vibration energy transfer in the collinear collision which approximates the He-H, system, The results are compared with the quantum calculations and several classical and semiclassical approaches. Very good agreement with quantum theory is found, forall the parameters investigated. PA 34.10.45; 34.50.Ez 1 Introduction Despite its drawbacks classical theory is often used for describing atomic and molecular systems [1-10]. The reason is that it is often much easier to implement than ‘quantum theory, for example, in the collision problems of atoms and molecules. However, there is no unique pre- scription how to use classical theory, because one sccks a compromise between what is thought to represent and what the sensible answer is. The assumption is that classi- cal theory answers the question: Given initial position and velocity of particle what are the values of these two quant- ities after certain time?, while the answers which make sense are: What the possible energies of bound particles are, or what the collision cross sections or the transition probabilities are? One can illustrate various uses of classical theory by analyzing a specific example, for instance, a collinear collision of an atom on a diatomic molecule, the later being represented by harmonic oscillator. The initial para- meters which determine this collision are the total energy of the system and the vibration energy of the molecule. After collision we expect to obtain probability of finding the molecule having particular vibration energy. Classical approaches vary according to how this information is extracted, but in all of them the common feature is the assumption that the initial relative position of the two atoms in the molecule is uncertain. The simplest approach is to use canonical variables: phase and enerey of the oscillator. The phase is assumed to be uniformly distrib- uted in the interval [0,2], while the energy is known precisely. It can be shown that this choice of the initial conditions is equivalent to choosing relative separation of the two atoms at random from the distribution which is, proportional to the inverse of the relative velocity of the two atoms if they are separated by this distance [11-13]. Transition probabilities are calculated by counting trajec- tories which leave the molecule in a vicinity of a given vibration energy. The exact procedure, however, would be to calculate the transition probabilities from the Jacobian, [14], ‘Classical results are improved by using the semiclassi- cal approximation, which is based on the Feynman's ‘method of the path integrals. The basic idea in this ap- proximation is to select only those trajectories which give the largest contribution to the transition amplitude in the limit h +0 [15-17]. In this way the interference effects are taken into account (and if dynamics is defined for complex, coordinates and time then even the tunneling effect can be described [18)). In all the other respects the semiclassical approximation is the same as the classical approach pre- viously described, This particularly concerns the choice of the initial conditions The above was a typical way how classical theory is, applied to atomic and molecular systems, and our inten- tion is to describe another one. We start by amending the basie principles of classical theory, and from that develop a self-consistent theory leading to the final result, the transition probabilities. The amendment is made by te quiring that classical theory answers the question: Given initial probability densities of finding particle at certain, position and with certain velocity, what are these prob- ability densities after certain time? Besides one requires that the two probability densities are related in such a way so that the uncertainty principle is satisfied. The question, is now: Is it possible to formulate classical theory with this, amendment [19], without any additional physical as- sumptions which Would somehow introduce quantum fea- tures into it? The uncertainty principle itself appears to introduce this feature, however, it should be stressed that, this is the law of nature about the existence of which we 108 earned from quantum theory, and its formulation does not immediately imply the underlying dynamics in nature. As it will be shown, we can indeed formulate classical theory with this amendment and this constraint. 2 Classical dynamic equation By postulating the uncertainty principle it is not self evident what the relationship. between the probability densities P(x) (in a one dimensional case) for the coordi- nate and Q(p) for the momentum should be so that it is satisfied. One can think of various ways to relate P(x) and (p), but the one which is well known assumes that we write Ped=lWedl OP) = ld)? @ where the two auxiliary functions and ¢ are related by vO) = rm In this way we answered a purely mathematical question, and still the dynamics equations are required which propagate y in time. Setting aside for the moment the question of dynamics, it should be mentioned that the quantitics which are being calculated have always average values. For example, the total energy of particle, in a po- tential V, has only meaning as the average quantity, given by |p o(p)e”™". Q) = Jape V6) PQ 0 of quantum theory it was done quite the contrary: Quantum theory was derived from classical, and the only step which sets the two theories apart was the assumption (10). Hence, when one talks about the differences between classical and quantum theory one essentially means the impact of this approximation on the classical results. For a particle which is bound the probability distribu- tion P(x, ¢) should be stationary. The eq (12) reduces then to the form Bodo 2m dx? where we assumed harmonic oscillator for the potential, and wo is its characteristic angular frequency. The sol- tions of (13) are familiar: the eigenergies are Ey = woln + 1/2) (a) where n =0, 1,2, ols) = NH, (ymin ie" ™r28 (as) where N is the normalization constant. Therefore, classi- cal theory (this term can be used because for harmonic ‘oscillator (13) is derived exactly from the classical prin- ciples) predicts infinite number of discrete probability distributions which are stationary. The appropriate phase space densities (7) are given by oe Pie Ehvo tj mw8? Ho (3) , While the appropriate functions Yo un 1 =L eure Pi == Quh — De (2u?/h? — dulh + 1em a9, Ft mast w ‘These densities will be used later in the study of atom- molecule collisions. 3 Calculation of transition probabilities In this section we will use previous results to calculate the transition probabilities in the collinear collision of an atom with a molecule. Internal coordinate of the molecule is designated by x, while the relative separation of the incoming atom and the center of mass of the molecule is y, Initially, say at t = 0, the atom and the molecule are well separated, the molecule being in the bound state mo, de- sotibed by the phase space density pay(x, p), and the atom is localized around some initial position, its probability density given by P,()). The overall probability density is o(x, ») and it is interpreted as the probability density of 110 finding the atom at the distance y and the molecule being stretched by x. Given the initial distribution we will now describe how to calculate the probability distribution P(x, y,t) at some later time, There are various ways to doing this, but the simplest is the following: two pairs of random numbers [pi] and [y;, q:] are generated, with the property that xa pel are distributed according to the phase space den- Sity Pag(x, p) While y, are distributed according to P,(y) and g, according to Q4(q). These numbers are the initial conditions for a trajectory, and after time r the value of the coordinates is x and y. It N sets of the initial conditions are generated, out of which n trajectories end in a small interval around the coordinates x and y, then the classical probability density at time ¢ is P(x, ys) = (18) Na, where 5, and 6, are intervals within which trajectories are sampled. An important observation at this point is neces- sary. The phase space density p(x, p) is not a positive definite quantity, the price we pay when the uncertainty principle was introduced into classical theory. In addition, this may be the sign that the Wigner’s function is not the best choice for the phase space density, and if another is found then this problem might disappear. However, since this function is mathematically related to the original premises of the theory, the problem of having negative values is just academic. The important physical quantities are derived from it, and as long as they are acceptable the question of the nature of p does not arises. One only has to use the phase space density properly and there is no reason that meaningless results should be obtained. The prescription how to use it follows from the relationship of to the meaningful quantities, e. the probability distri- bution P(x), given by (5) In general, if'a quantity F(x) is represented as P(x) = fap ple p) f(p) (19) then the integral is evaluated by the Monte-Carlo method from the transformed form F(x) = fdpip (x, pilSignl p(x, PL FP) (20) where the module of p represents a positive definite quantity which is used as the distribution in the Monte- Carlo method, and the sign of p which multiplies f(p) is the integrand. The prescription for the calculation of P(x) is, therefore, the following: Random numbers are gener- ated from [p(x, p), and if the pair [x,, pi] is such that p(x, p)) >0 then +1 is associated with the trajectory, while if p(x, p:) <0 then —1. The number 7 in (18) is now obtained by subtracting the number of trajectories which hhaye —1 associated with them from the number of trajec- tories which have +1. This method is only a more general use of the Monte-Carlo method when the probability distribution is not positive definite, but mathematically speaking itis well defined. ‘The meaning of P(x, y,t) is obvious: It gives the prob- ability density at time ¢ of finding the atom at the distance y from the molecule, if at the same time the molecule is Stretched by x. The question is now how to obtain mean- ingful information from this quantity, in the first place the probability of finding the molecule in the state n. These transition probabilities are easily obtained if the function WGsy, 1) is known, which has for a collinear collision problem the most general form Ynol% Jot) = Sdk AK) bo (% Ys Rem @y where the index mp designates the initial state of the molecule, The function dya(% y. K) isa solution of the time independent Sehrédingers equation, subject to the scat- tering boundary condition bral Ys K) = Wool) eM + Fi dasnolk)Walspe“%™ (22) ‘where wa(x) are eigenfunctions of the molecule (for simpli- city it is assumed that the molecule cannot dissociate, however, generalization to include this possibility is straightforward). The coefficient A(k) is determined from the initial y for the ineoming atom, After a long enough time, when the entire incident probability distribution is scattered, and it is far away from the scattering region, the function is in the form Wael ¥, 8) = LwWal) [dk A(K) dy, go (Klee (23) By definition the coefficient of w,(x) is the transition amplitude into the state n, and its modulus squared is the required transition probability. Therefore, if y/(x,y, 0) is Known (the indices sc and ng are omitted for convenience) the transition amplitudes are easily obtained. From classi- cal theory, however, one calculates P(x, y,¢}, and. this gives information only about the modulus of w(x, 9, ¢) namely Poy d=» OF (24) while the phase of ye, ) is still undetermined, It can be calculated by writing 65,950 =1¥la x plea es and from the probability current in the coordinate x ses. n0)~ Rim yes 3.9 CEE west) =" Poy 90) Beans i 026) we obtain information about the derivative of the phase. Equation (26) has solution for the phase (up to a constant) H's ys) PRL YD where the current j is calculated from classical theory. This, however, implies that the entire function y is also calculated from classical quantities. ‘The current jis calculated in the same way as P(x, y, ®), however, instead of adding +1 (depending on the sign of the phase space density p) for each trajectory which ends {in a small interval 6, and 5, around the final coordinates x and y, respectively, one adds the value of the final velocity +tv(x, yf) (based on the evaluation of the integ- rand in (19) where f(p) = p/m = 0). In principle, therefore, 50x, Wat fax’ en th di ve ity tu ar bh at Inge ns 2 be 22a a ti ———————— the phase 6(x, y,2) can be calculated but the numerical difficulties may be encountered in the region where adding velocities results in the cancellations because the probabil- ity current is too small This procedure for calculating the transition ampli- tudes is quite straightforward but numerically very de- ‘manding. There is, however, an alternative procedure which requires less numerical effort, and it consists of calculating the average probability density Pls) = fdyPlx y.0) (28) meaning that the position of the atom after collision is immaterial. Calculation of (28) is simple: the trajectories are counted in the interval around x irrespective of what the value of y is. ‘The average probability density (28) can also be cal- culated from quantum theory. Starting from (23) the prob- ability density is (24), and if it is integrated over the coordinate y one obtains Px) = Swalsdwe od fd fdk’ A) Ark) xe MEE, KS wl K)226 (hy Ks) (29) and after integration over k’ we get ey a eH dy alk AE (Kase) 0) where the nonessential factors were omitted. A new vari- able was introduced, defined by kyw = J? — xi +12, where Kz = 2yeq/h® and ys is the reduced mass of the system. Energy of the bound state m is &,. If in addition P(x, 1) is averaged over time Ff aeons) G1) Pox) L wald ww (x) [dk A(K)A® ke) Pia) where T is large, then 2 Fagg Pls) = we fake 1A) [eno (32) ‘The amplitude A(k) is approximately the delta function in the limit of the incident plane wave with the dominant contribution from the initial momentum k = pp. Finally we have POs) Dw 0) Paw 63) where Py. ate the required transition probabilities, They can be calculated by making the least-square fit of the classical P(x), when (28) is time averaged. In our particu- Tar example, w,(x) are the eigenfunctions of the harmonic oscillator (15). 4 Examples and discussion A system was analyzed which mimics He-H collision. In the center of mass of the three atoms the classical equa- tions of motion are tu GV (y ~ x/2) a ee 4 VY =x?) xt 64 where M is mass of the incoming atom and m is mass of one of the atoms in the homonuclear diatomic molecule. The potential of the diatomic molecule is assumed to be harmonic oscillator, of the form «x?/2, while the potential between the incoming atom and the diatomic molecule is, Voexp[— ay/A(2y — x)], where 4 = mw (2A) [45] The choice of the initial conditions for the internal coordinate and momentum of the molecule has been de- scribed earlier. For the incoming atom, on the other hand, it is assumed to be uniformly delocalized in a large inter- val for y < yo, where yo is sufficiently far from the mole- cule so'that the potential V is negligible, The interval is sufficiently large so that the relative momentum of the atom and the molecule has a fixed value, and given from the energy conservation. In practice the semi-infinite ex- tent of the interval is ensured by limiting it to length 2nVo/w, where Vo is the relative velocity of the atom and the molecule. This is because the final value of x is periodic with this interval. In the first example we calculated transition probabil ities when the molecule is initially in the ground state (1 =0). The averaged probability distribution (33) was calculated for time t which was long enough so that all the ‘chosen trajectories are scattered and the test was made to check that the averaged probability distribution is inde- pendent of the time when the sampling is done. Two examples of these probability distributions are shown in Fig. 1, one for the low and the other for the high collision energy (collision energy is measured in the units of hw/2, jie. the zero point energy of the harmonic oscillator). The circles represent classical calculation based on 90000 sets of the initial conditions, while the solid line is the “es ; 7 ca a a, Fig. 1. Typical east-square fit (solid Ite of the time-averaged classi- cal probability distribution (circles) when the molecule is intially in its around state 12 ‘Table L. Transition probabilities from initial ground state of Ht, molecule for B= 6¢h0/2) E1002) E=120hw?2) = 204hw/2) ‘various collision energies. Comparison is aaa sade with the quantum calculations of Pron Pm Pam Paw Pare Prima Pam Sssrest and Johnson © 0988 0978 O74) «0733-0527 «0838005 0060, 1 0016 00220247 02520400304 0.202 O28, 2 OO15 00150064 106803520366, 3 0294 0267 2 0.107 0092 10000 ‘at 000000 Net Net : 8 " Fig.2. Transition probabilities in He-H, collision when the molecu Jeis initially i its ground vibrational state, Quantum (slid line) and classical (broken line) calculations are shown for four typical coll- sion energies. Solid triangles are results ofthe conventional clasical theory while the solid squares are that of the semiclassical theory, ‘based on the path integral method least-square fit of the form (33), where the number of the basis functions was chosen in order to ensure that the results converged. It is interesting to note that no large ‘numerical instabilities were encountered in the least- square fit procedure, despite the fact that the basis fune- tions are all positive definite. The transition probabilities were calculated for 4 dif- ferent collision energies, and the numerical values of the classical and the quantum calculations [45] are given in Table 1, The results are shown in Fig. 2, where the solid line connects the quantum transition probabilities, while the broken line are the results of the classical calculation, obtained from the least-square fit (33). The triangles show results of the conventional classical calculation [15], and the squares are that of the semiclassical calculation based ‘on the Feynman's method of path integrals [15]. At lower energy these results are not shown because their agree- ‘ment with the quantum results is very bad. We notice that the results of the semiclassical calculation do not show teat improvement over the conventional classical results. ‘The reason for this was explained in Introduction, where it is argued that the semiclassical theory does not go beyond the basic principles of the conventional classical theory, except that the interference effects are taken into account. On the other hand, agreement between our iM ee 2 10% oP xm Fig.3. Typical least square fit (solid Ine) ofthe time-averaged clasi- cal probability distribution (circles) when the molecule is intially in the frst excited stat. The wo figures show results of two choices for, ‘the number of sets of initial conditions zeal 2 oy cam Fea . Fig. 4. The same as in Fig. 2 except that the molecule is intially in the fist excited vibrational state classical and quantum calculations is very good, becom- ing progressively worse as the collision energy increases. Similar results are obtained when the molecule is ini- tially in the excited states. Figure 3 shows a typical aver- aged probability distribution (33) when the molecule is initially in the first excited state. The two distributions Tat firs) Sec Tak eo Fig cal wer tior sent dist abil (ol agr beir trar mol trib the tar ities give ‘Table2. Transition probabilities from the first excited state of H, molecule for z various collision energies. Comparison is ‘made with the quantum calculations of Pane Secrest and Johnson ‘0012 0022 0999 0977 001 0.001 ‘Table3. Transition probabilities from Initial state of H molecule for ‘various collision energies. Comparison is made with the quantum eaiculations of Secrest and Johnson o 1 2 3 4 3 6 Ge a3 a3 8 a 0b ose Fig. 5. Typical least square fit solid line) of the time-averaged classi- cal probability distribution (circies) when the molecule is intially in were calculated for 10000 and 10° sets of initial condi- tions, in order to make sure that the final results are not sensitive to this number. The least-square fit of the two distributions produced virtually the same transition prob- abilities. Figure 4 shows comparison of the quantum (Golid line) and our classical results (broken line). The agreement is again very good, for high energy, though, being less so, Table 2 gives the numerical values of the ‘transition probabilities. Finally, examples with the initial n= 2 state of the ‘molecule were analyzed. Typical averaged probability dis- tributions are shown in Pig, 5. Solid line represents now the sum (33) where the coefficients are the quantum transition probabilities. The relevant transition probabil- ities are shown in Fig. 6 and their numerical values are given in Table 3. The agreement is as good as in the Fait) Ea iin B= TOO Pr Pras O36) 022 aoe ome 0306 ane Sie Seq fee ORs OR Oe mee ise ase OMS Onn ns <0310 Omi Sole 8857 ome alo, aa Boao oi aan ‘12(hw/2) E = Wihw/2) Pore Pea Pete Praaet 0455 0058 0366 036 Ome ses Gost aan bmp O38 Gis 37 oane Om ids ae =O618 fone ash ale Oey Ole Oso Gos? & ef SE Fig. 6. The same as in Fig, 2 except thatthe molecule is intially in n= 2 vibrational state previous cases, from which one anticipates that the same ‘would be expected when the molecule is initially in even higher states. ‘Based on the previous results one concludes that classi- cal theory gives almost the same results as the quantum, the deviations being relatively large for the higher collision energies. The cause of the deviations is not clear, and 1 possible explanation is that the unharmonic forces, ori- inating in the repulsive potential between the atom and the molecule, become progressively more important as the collision energy increases. In such a case the term which was neglected in (10) play a significant role and the close connection between classical and quantum theory is lost. However, more detailed analysis of this point is required. We have also demonstrated some of the more funda- ‘mental aspects of this classical approach, One of the major a na results that bound state, including the excited ones, are easily incorporated into the classical calculations. In fact wwe have shown that the concept of bound states is entirely classical, once the uncertainty principie is incorporated into classical theory. Not only that classical theory, in the form presented earlier, gives discrete stationary states, but one can also get meaningful results from them. Various cited examples ofthe use of this amended classical theory confirms that this approach to classical theory is well founded. References Rutherford, EB: Phil. Mag. 21, 669 (1911) ‘Abrines, R., Percival, LC: Proc. Phys. Soc. 8, 861 (1966) Cohen, 1S. Phys. Rev. A36, 2024 (1987) Cocke, C.L: Phys. Rev. A20, 749 (1979) Talley, KC, Preston, RK: I. Chem. Phys. $5, 562 (1971), Yinnon, A'T, Bosanac, S, Gerber, RB, Murrell, JN: Chem. Phys, Lett. 58, 364 (1978) 7, Karplus, M, Porter, RN, Sharma, R.D: 1. Chem, Phys. 43, 3259 (1965) Bosanac, SD: Phys. Rev. A22, 2617 (1980) Benson, SW, Berend, G.C, Wa, J.C: J. Chem. Phys. 38, 25, (1963) 10. Porter, RIN, Raff, LM: In: Dynamics of molecular collisions Miller, W-H. (ed), Vol. B, p. 1. New York: Plenum Press 1976 ‘This conclusion fests on a more general statistical approach ‘based on the use of microcanonical distribution in the phase space, where what we have just said is described by the delta function distribution, ee [2] 12 Bohm, D: Quantum theory. Prentice 1952 1B, Berry, MLV, Mount, KE: Rep. Prog, Phys. 36, 315 (1972) 14 Murrell, LN, Bosanac, SD. Introduction to the theory of atomic and molecular collisions, p. 84. New York: Wiley 1989 15, Miller, WAL: Adv. Chem. Phys. 25, 69 (1974) eee 16. 17 18 19. 20, 21 2 3 24 26, 2 2 2, 20. 31 32 33. 34 35. 36 37 38 3. 41 2, 4 “4 4s, Miller, W.HL: J. Chem. Phys. 53, 3578 (1970) Feynman, RP, Hibbs, AR; Quantum mechanics and path inte arals, New York McGraw-Hill 1964 Miller, W-H., George, T-F: J. Chem. Phys. 56, $668 (1972) Introducing the Heiseaberg’s uncertainty principle into classical theory has been suggested before (Kirschbaum, C.L., Wiles, L Phys. Rev. A21, 834 (1980), however, it was done in an ad hoe ‘manner by limiting artifially the range of definitions of the coordinate and momentum of particles Bosanae, SD: J. Phys. A27, 1723 (1994) Bosanac, S.D: Phys. Rev. ASO, 2899 (1994) Bosanac, SD: J. Chem, Phys. 95, 5732 (1991) Bosanac, SD: Z Phys. D28, 195 (1993) Bosanac, S.D: Z. Phys. D24, 325 (1992) Bosanae, S.D: Z Phys. A26, 3523 (1993) Doslic, N, Bosanae, S.D.Z. Phys. D32, 261 (1995) Bosanae, &D., Doslc, N: Fizika 93, 175 (1994) ‘Wigner, E: Phys. Rev. 40, 749 (1932) Hillery, M,, O'Connel, RF, Scully, M.O, Wigner, E.P: Phys Rep. 106, 122 (1988) Carruthers, P. Zachariasen, F: Rev. Mod. Phys. $5,245 (1983) Heller, EJ: J. Chem, Phys. 67, 3339 (1977) Berry, MLVe Phyl. Trans. Roy. Soc. 287, 237 (1977) Das Gupta, S: Nucl, Phys. A47H, 417e (1987) Suruad, E, Pl, M, Schuck, P: Nucl. Phys. A482, 187¢ (1988) a Providécia, J. Je: Nucl. Phys. A495, 193 (1989) Eichenauer, D, Grin, N, Scheid, We J. Phys, B14, 3929 (981) Horbatseh, M, Dreier, RM: Phys. Lett, A113, 251 (1985) Horbatsch, M: J. Phys. B25, 3797 (1992) Horbatsch, M: Phys. Rev. Ad9, 4556 (1994) Les, J, Soil, He J. Chem, Phys. 73, 238 (1980) Schinke, Rs Collision theory for atoms and molecule, Giant- treo, FAA. (ed), p. 241. New York: Plenum 1989 Jans, W., Moateiro, TS, Schweizer, W., Dando, PA: J. Phys. 'A26, 3187 (1993). Voros, A: Lect. Notes Phys. 93, 326 (1979) ‘Tabor, M-: Adv. Chem. Phys. 46, 73 (1981) Secres, D, Johnson, BR: J. Chem. Phys. 45, 4556 (1966) al 'D al tie tit ch in ot tw a nm ti

Você também pode gostar