Because of their experimental accessibility, the reactions of F and Cl with H2 and its isotopomers have become paradigms for triatomic abstraction reactions.[1,2] Experimental interest in the Cl + H2 reaction goes back more than 100 years.[1] Numerous quasiclassical trajectory and various quantum scattering investigations have been reported[3–6] on several potential energy surfaces (PESs).[7–9]
Approach of the H2 molecule to a Cl atom splits the degeneracy of the 2P state. As shown schematically in Fig. 1, two electronic states (12A′ and 12A″; 2Σ1/2+ and 2Π3/2 in linear geometry) correlate adiabatically with the ground-state atomic reactant (2P3/2) while a third state (22A′; 2Π1/2 in linear geometry) correlates adiabatically with the excited-state atomic reactant (2P1/2).[10] Of these, only the 12A′ electronic state correlates with the electronic ground state of the products [HCl(X 1Σ+) + H(2S)]. The two other electronic states correlate with electronically excited states of the products [HCl(a 3Π) + H(2S)] which are considerably higher in energy[11] and, consequently, energetically inaccessible at low to moderate collision energies.
It is common to assume that the motion of the atoms in a chemical reaction occurs on a single potential energy surface (PES). This description is a consequence of the Born–Oppenheimer (BO) approximation. For bimolecular reactive collisions this assumption is supported by the majority of past experimental work.[12] In recent molecular beam experiments, Liu and co-workers[13] used two different Cl atom sources to separate the reactivity of the two SO states of the Cl atom. Strikingly, except at the lowest collision energies, they conclude that the excited SO state has a significantly larger reactive cross section. As can be seen in Fig. 1, reaction of the spin–orbit excited Cl atom (2P1/2) can occur only by a non-Born–Oppenheimer transition from the non-reactive 2Π1/2 PES to the reactive 2Σ+1/2 PES. When the two PESs are widely separated, there is little likelihood that this transition can occur. Hence, non-adiabaticity in the Cl + H2 reaction, possibly to the extent suggested by the experiments of Liu and co-workers, must be a result of couplings between the various PESs in the entrance channel of the reaction.
Theoretical investigations of the role of excited electronic surfaces in the F + H2 reaction date back to the pioneering work of Tully.[14] Several years ago, we presented the framework for the essentially exact quantum determination of cross sections for abstraction reactions involving an atom in a 2P electronic state.[15] This treatment involves six three-dimensional hypersurfaces, four of which describe the diabatic potential energy functions and two of which describe the coordinate dependence of the spin–orbit coupling. All Coriolis coupling terms are included.
More recently, we applied[5] this methodology to the reaction of Cl with molecular hydrogen, using new ab initio potential energy surfaces developed by Capecchi and Werner (the CW PESs).[9] Calculated integral cross sections for the non-Born–Oppenheimer reaction of the spin–orbit excited atom, which are a measure of the degree of non-adiabaticity in the reaction, were found to be markedly smaller than those for the Born–Oppenheimer allowed reaction of the ground spin–orbit state.[5] The degree of non-adiabaticity was even less than found in our earlier investigation of the F + H2 reaction.[15,16] This is not surprising, since the spin–orbit splitting in the Cl atom (882.35 cm–1 = 2.528 kcal mol–1) is more than twice as large as that of the F atom (404 cm–1).[17] Only at the lowest collision energies, where the ground-state channel is suppressed by the high barrier (Fig. 1), will reaction of the spin–orbit excited atom dominate. These theoretical results stand in direct contrast to the conclusions of Liu and co-workers.[13]
This past year[18] we have determined differential cross sections for the Cl + H2 reaction.[6] Differential cross sections for the adiabatically-forbidden reaction of Cl* (2P1/2) are backward peaked, similar to those for reaction of the ground spin-obit state, but are much smaller in magnitude.
Because of the continuing disagreement between experiment[13] and theory,[5] it is worthwhile to examine in more detail the factors which control the degree of non-adiabaticity in the Cl + H2 reaction. Coupling between the adiabatically allowed and forbidden pathways is due to (i) the spin–orbit Hamiltonian, (ii) electronic mixing of the two states of A′ symmetry, and (iii) Coriolis coupling between the rotational motion of the nuclei and the spin and electronic-orbital angular momenta of the Cl atom. One great advantage of a theoretical simulation, as compared to experiment, is that each of these terms may be altered (or eliminated) in turn to allow their effect to be examined. This is the goal of the present article.
Hamiltonian, potential surfaces, and couplingsWe write the total Hamiltonian for collision of Cl (2P) with H2 (or any other diatomic) molecule as H(R⃑,r⃑,q⃑) = Tnuc(R⃑,r⃑) + Hel(q⃑;R⃑,r⃑) + Hso(q⃑;R⃑,r⃑).Here q⃑ is a collective notation for the electronic coordinates, Hel is the electronic Hamiltonian, which, in the Born–Oppenheimer sense, depends parametrically on the positions of the three nuclei, and Hso is the spin–orbit Hamiltonian, which is not included in Hel. The nuclear coordinates R⃑ and r⃑ in eqn. (1) designate the Jacobi vectors[19] in any one of the three chemical arrangements (Cl + HH, H + HCl, or HCl + H). In terms of the mass-scaled Jacobi coordinates S and s, one defines Delves hyperspherical coordinates ρ and θ.[19]
At each value of the hyperradius ρ we expand the total wavefunction in an overcomplete basis of rotational-vibrational-electronic wavefunctions in each arrangement. In the Cl + H2 arrangement there are six electronic states. For a halogen atom with a p5 electron occupancy these states correspond to the three spatial orientations of the p hole and the two possible spin-projection quantum numbers. For the HCl, the lowest excited state is of Π symmetry, and lies far above the ground state (X 1Σ+) in the region of the molecular minimum.[11] If we neglect these excited Π states, then, in each H + HX arrangement, we need retain only two states, which correspond to the HX molecule in its ground electronic state (1Σ+) combined with the two possible spin-projection quantum numbers of the H atom.
In each arrangement the basis functions are[15,20,21] Here J is the total angular momentum, with projection M along the space-frame z-axis and projection K along the Jacobi vector R⃑ of the relevant arrangement. The quantum number j designates the rotational angular momentum of the diatomic moiety in the same arrangement, with projection k along R⃑. Here, also, DJ*MK(Ω) is a Wigner rotation matrix element,[22]Yjk(γ,0) is a spherical harmonic,[22] and φvj(θ;ρ) is the solution of a “vibrational” reference problem corresponding to motion in the hyperangle θ at the given (fixed) value of ρ.[23]
The quantities λ and σ in eqn. (2) are the projection of the electronic orbital and spin angular momenta along R⃑. Since we consider here only doublet electronic states, the spin part of the wavefunction corresponds to s = 1/2 and σ = ±1/2. The projection quantum numbers are related as follows K = k + λ + σ.To solve the close-coupled reactive scattering equations, one must first construct surface functions in each sector,[20,24] by diagonalizing the total Hamiltonian, exclusive of the kinetic operator corresponding to radial motion in the hyperradius ρ, in the multiple arrangement basis consisting of the states defined by eqn. (2). In addition to the electrostatic interaction potential and the spin–orbit Hamiltonian, it is also necessary to determine matrix elements of the orbital angular momentum L2 = (J – l – s – j)2.
In evaluating the matrix elements of the electrostatic interaction potential one first integrates over the electronic coordinates to obtain formally 〈λ′σ′|Hel(q⃑;R⃑,r⃑)|λσ〉 = δσσ′Vλ′λ(ρ, θ, γ). Similarly, the matrix elements of the spin–orbit interaction potential involve the initial determination of the electronic matrix elements 〈λ′σ′|Hso(q⃑;R⃑,r⃑)|λσ〉 = Wλ′σ′λσ(ρ,θ,γ). Once these matrix elements, which are functions of the three internal coordinates, are evaluated, the scattering calculations are no different from that of a triatomic reaction on a single potential energy surface.
Capecchi and Werner[9] used internally-contracted multi-reference configuration-interaction (IC-MRCI) calculations, based on state-averaged (three-state) complete active space self-consistent field (CASSCF) eference wavefunctions with very large atomic basis sets, to determine the three electronically adiabatic Cl + H2 PESs in the reactant arrangement: 1A′, 2A′, and 1A″. These all correlate with Cl(2P) + H2. These three adiabatic electronic states are the IC-MRCI approximations to the three lowest eigenfunctions of Hel, namely Hel(R⃑,r⃑;q⃑)ψi(R⃑,r⃑;q⃑) = Ei(R⃑,r⃑)ψi(R⃑,r⃑;q⃑), where the subscript i = 1,2,3 designates the |1A′〉, |2A′〉, and |1A″〉 states. The coordinate dependence of the eigenvalues Ei defines the three adiabatic PESs. We define the interaction potential by subtracting the energies of the reactants at infinite Cl–H2 separation, so that Vi(R⃑,r⃑) = Ei(R⃑,r⃑) – ECl – EH2(re).Here ECl designates the electronic energy of the Cl atom in its 2P state exclusive of the spin–orbit Hamiltonian. The zero of energy will be defined by the equilibrium internuclear separation of the H2 molecule as determined in the IC-MRCI calculations, so that Vi(|R⃑| = ∞, |r⃑| = re) = 0,where re is the equilibrium internuclear separation of the H2 molecule.
By analysis of the coefficients in the CI expansion of the ClH2 electronic wavefunctions, the two PES's of A′ reflection symmetry are transformed to an approximate diabatic basis,[25,26] in which the orientation of the p hole on the Cl atom is fixed with respect to the plane defined by the three atoms. Here we take R⃑ to define the z-axis and chose the y-axis to be perpendicular to the triatomic plane. We shall designate the diabatic states by the projections of the electronic orbital and spin angular momenta along R⃑. We use the compact Cartesian notation |Πx〉, |Πy〉, and |Σ〉. The two adiabatic states of A′ symmetry correspond to a 2 × 2 rotation of the |Πx〉 and |Σ〉 diabatic states, namely whereHere the transformation angle ζ depends on R, θ and γ (or, equivalently, ρ, s and γ. Since there is no coupling between the |Πy〉 state, of A″ reflection symmetry, with the |Πx〉 and |Σ〉 states, of A′ reflection symmetry, the adiabatic and diabatic states of A″ reflection symmetry are identical. Alternatively, we can define the diabatic states in terms of signed-λ, rather than Cartesian, projections, namely |Π1〉, |Π–1〉, and |Σ〉. These signed –λ states are those that appear above in the expansion of the scattering wavefunction. Note that the state we designate as “Σ” corresponds to λ = 0.
As presented in detail in an earlier paper,[15] the matrix of the interaction potential, in the 6 × 6 basis defined by the three Cartesian diabatic states and the two possible spin projections (which are also defined with respect to R⃑) can be described in terms of three diagonal, electronically diabatic PES's: Vxx, Vyy, and Vzz,[27] as well as a fourth PES, Vxz, which is the coupling between the two states of A′ symmetry. Each of these four PES's is a function of the three internal coordinates. Thus, the description of the interaction in the electronically diabatic basis, which involves these four potential energy functions, is equivalent to the description in the electronically adiabatic basis, which involves three potential energy surfaces (V1A′, V2A′, and V1A″) plus the coordinate dependent mixing angle ζ.
In the basis of the signed-λσ states, which are used in eqn. (2) in the expansion of the scattering wavefunction, the matrix of the interaction potential, which we represented schematically as Vλ′λ(ρ,θ,γ) in eqn. (4), is Here, a bar over the state label designates a spin projection of –1/2 (β spin). The relation between the matrix elements in eqn. (11), and those determined by the ab initio calculations, which are carried out in a Cartesian basis, are VΣ = Vzz, VΠ = (Vyy + Vxx)/2, V2 = (Vyy – Vxx)/2, and V1 = Vxz/21/2. In collinear geometry both V1 and V2 vanish. In perpendicular geometry V1 also vanishes while V2 goes through a relative minimum. Figs. (2) and (3) display contour plots of the four diabatic PES's (VΣ, VΠ, V1 and V2).
Note that the ordering of the states is here grouped by values of ω = λ + σ, and is consequently different form that chosen in our earlier paper.[15] As seen in Fig. 1, four states correlate with the Cl(2P3/2) + H2 reactant (ω = ±1/2 and ω = ±3/2). Only two states correlate with the Cl(2P1/2) + H2 reactant (ω = ±1/2).
The matrix of the spin–orbit Hamiltonian is determined fully by two components,[26]A(R, r, θ) ≡ i〈Πy|Hso|Πx〉 and B(R, r, θ) ≡ 〈x|Hso|Σ〉, where The spin–orbit functions A and B can be determined in the ab initio calculations, along with the PESs. The dependence of these on the reactant arrangement Jacobi coordinates are shown in Fig. 4. In terms of A and B, the matrix of the spin–orbit Hamiltonian in the |λσ〉 basis is If the V1 and V2 terms are neglected, then the electrostatic Hamiltonian [eqn. (11)] is diagonal in the λσ basis while the matrix of the spin–orbit Hamiltonian [eqn. (15)] is blocked into two identical 2 × 2 matrices and two single channels.
Consequently, the matrix of Hel + Hso in the ω = 1/2 (or equivalently, ω = –1/2) states, is At long, range, where VΣ and VΠ vanish, this can be diagonalized by transforming to a coupled (jaω) basis (where ja is the total electronic angular momentum of the molecule), with energies –A/2 for the ja = 3/2 states and +A for the ja = 1/2 states. As short range, where the splitting between VΠ and VΣ is much larger than the spin–orbit constant, the Hamiltonian is approximately diagonal in the (uncoupled) λσ basis.
Early work on inelastic collisions of atoms in 2P states has shown[28] that non-adiabatic transitions will take place predominately at the value of R where the difference between the diagonal terms in eqn. (16), which increases as the collision partners approach, becomes equal to the off-diagonal term. At this point a frame transformation takes place. This point (or two-dimensional seam in the three-dimensional ClH2 PES) is marked by a heavy line in Fig. 2. It is in the neighborhood of this seam that the nonadiabatic transitions will occur, mixing the ω = ±1/2 states which correlate with Cl(2P1/2) and the ω = ±1/2 states which correlate with Cl(2P3/2). We observe that this transition point occurs well outside the barrier to reaction.
Further, we observe from Fig. 3 that in the region of this seam both the V1 and V2 coupling potentials are considerably smaller than the off-diagonal spin–orbit coupling (21/2B ≅ 3.57 kcal mol–1). Fig. 5 compares the off-diagonal coupling in eqn. (16), the splitting between the diagonal terms, (VΠ + A – VΣ), and V1, as a function of R for the hydrogen molecule held at its equilibrium distance (1.4 bohr) and a fixed Jacobi angle of γ = 45°.
Finally, we observe, by comparing Figs. 2 and 4, that in the region of the seam in Fig. 2, where, presumably, nonadiabatic transitions are most likely to occur, the two spin–orbit terms are little changed from their asymptotic value.
We see from matrices (11) and (16) that both the spin–orbit and electrostatic Hamiltonians can cause transitions between the spin–orbit excited reactant channel to the ω = ±1/2 and ±3/2 states which correlate with the ground spin–orbit reactant channel. In addition, off-diagonal Coriolis coupling, which arises from the expansion of the orbital (end-over-end) angular momentum operator for the relative (end-over-end) motion of Cl with respect to H2, can also contribute to spin–orbit changing transitions.
Since L = (J – l – s – j), one can show, that the effect of L2 on the |JMKvjkλσ〉 states of eqn. (2) is where, for simplicity, we have suppressed the J,M,v,j quantum numbers. Here, we define a± ≡ 〈λ ± 1|l±|λ〉.
Premultiplying by various 〈JMK′v′j′k′λ′σ′| states will give the matrix of the rotational Hamiltonian. The first term in eqn. (17) is responsible for the diagonal matrix elements. The second, third and fourth terms, which all vary roughly linearly with J, will make a more significant contribution than the other terms, which are independent of J. The former terms give rise to the matrix element This term is to be multiplied, of course, by 1/(2µR2), where µ is the Cl–H2 reduced mass (1.96 u). The third term in eqn. (18) is identical to the off-diagonal Coriolis term which appears in the standard centrifugally decoupled (coupled-states) treatment of atom-molecule collisions without electronic degrees of freedom.[29,30] Because of the δλ′λδσ′σ factor, this term will mix the two ω = 1/2 states which correlate with the two spin–orbit manifolds. In the region of the non-adiabatic mixing seam (Fig. 1), the magnitude of this term will be approximately 2.5 × 10–3jJ kcal mol–1. The other two terms will have magnitudes approximately 2.5 × 10–3J. Consequently, since the rotational angular momentum of the H2 molecule (j) will never be larger than 4 or 5, we predict that the contribution of the off-diagonal Coriolis coupling to nonadiabaticity in the Cl + H2 reaction will be important only for total angular momentum J greater than about 25.
As discussed in the introduction, theoretical simulation allows us to vary (or eliminate) one (or more) of these terms which contribute to the nonadiabaticity in the reaction and hence investigate the relative importance of these various terms. This will be explored in the next Section.
The time-independent, close-coupling calculations we use to investigate the scattering (both reactive and inelastic) of Cl by H2 have been described in detail in several earlier publications. For each value of the hyperradius ρ, we expand the total scattering wavefunction as outlined by eqn. (2). A log-derivative method[31] is used to propagate the solution numerically, from small ρ to large ρ. The parameters which control the accuracy of the integration are increased until the desired quantities (transition probabilities, integral and/or differential cross sections) have converged to within a reasonable limit. For abstraction reactions dominated by linear (or near linear) barriers, converged results can be obtained with only a few values of K, which greatly reduces the necessary computer time. We use the ABC code of Manolopoulos and co-workers,[23,32] extended to include the electronic degrees of freedom and multiple PESs.[6,15]
To reduce further the number of basis functions which must be simultaneously considered, we use definite-parity linear combinations of the signed-K rotation–vibration–electronic basis functions (eqn. (2)). Further reduction in the size of the basis can be made by using eigenfunctions of the H ↔ H interchange symmetry. The propagation of the solutions to the CC equations are carried out entirely in the fully-uncoupled, body-frame basis of eqn. (2). At the end of the propagation, where all the PESs (VΣ, VΠ, V1, and V2) are negligibly small, but before extraction of the S matrix, we transform the log-derivative matrix into a partially-coupled basis which diagonalizes the residual spin–orbit coupling [eqn. (16)]. In this partially-coupled basis, the electronic states, which correspond to the Russell–Saunders spin–orbit states of the Cl atom, are labeled by the total electronic angular momentum of the atom, ja, and the projection of ja along R⃑, which we designate ka.
Figs. 6 and 7 display the total probability of reaction (summed over all energetically accessible rotational and vibrational states of the HCl products) as a function of collision energy. These figures refer to reaction of H2 in v = 0, j = 0 (the lowest state of para-H2) at the lowest possible value of the total angular momentum J = 1/2. The results shown refer to states of positive overall parity; the results for the states of negative parity are virtually identical. Results are shown for calculations with the full Hamiltonian, including all couplings (spin–orbit, electrostatic, and Coriolis) as well as from calculations in which we (a) eliminated V1 (the coupling between the two states of A′ reflection symmetry), (b) eliminated V1 and V2 (the coupling between the two Π states), (c) held the two spin–orbit constants to their asymptotic value (A = B = 2.52 kcal mol–1), or (d) eliminated all off-diagonal Coriolis coupling.
The probabilities for reaction out of the ground (2P3/2) and excited (2P1/2) spin–orbit states are shown separately in Fig. 6, and together in Fig. 7. As we have remarked earlier,[5] the stair-case structure in the energy dependence of the probabilities for reaction of Cl(2P3/2) correspond, in a transition-state model, to the opening of successively higher bend-stretch states of the ClH2 complex at the transition states. The position of the lower of these states is shown in Fig. 1.
We observe that the results change little when the approximations discussed in the preceding paragraph are introduced. Accurate reaction probabilities can be obtained by retaining only the spin–orbit coupling, and, further, without taking into account the small variation of these constants with the approach of the H2 molecule. We might have anticipated this result from the discussion in the preceding Section. In the region where significant non-adiabatic transitions will occur, at the point where the off-diagonal spin–orbit coupling (21/2B) is equal to the splitting between the two ω = 1/2 states, the V1 and V2 coupling potentials are only a small fraction of this off-diagonal coupling. In reality, this relative size difference will be greater than is seen in the comparison of Figs. 3 and 5, because all Jacobi angles γ are sampled during the collision. Both V1 and V2 reach the maximum values shown in Fig. 3 at γ = 45° and decline sharply as γ goes to zero (where they both vanish) or to 90° (where V1 vanishes and where V2 goes through a local minimum).
We also observe that the off-diagonal Coriolis coupling contributes little to the nonadiabaticity. However, since three of the off-diagonal Coriolis terms increase roughly linearly with J, it may be worthwhile to investigate the effect of neglecting the Coriolis coupling at higher values of the angular momentum. Fig. 8 shows the dependence on the collision energy of the total reaction probabilities for several values of the total angular momentum. The total reaction probability decreases with increasing J. This is a consequence of the the increasingly large centrifugal barrier which shifts the threshold to higher energy.
To investigate the effect of increasing J on the validity of the approximations made earlier, we present in Fig. 9 plots of the total reaction probabilities out of H2(v = 0, j = 0), positive parity states at J = 17.5. Although the spread in calculated transitions probabilities, as a function of the various simplifying approximations, is a bit larger than seen in Fig. 6 at J = 0.5, nevertheless, we see that the conclusion made earlier for J = 0.5 is still valid, namely that the degree of inelasticity in the Cl + H2 reaction can be determined to a high degree of accuracy by neglecting all off-diagonal couplings other than the spin–orbit coupling.
As we have seen here, the overall degree of inelasticity in the Cl + H2 reaction is governed, almost exclusively, by the spin–orbit coupling. If the smaller V1 and V2 PESs, as well as Coriolis coupling, can be safely neglected, then the computational problem becomes considerably reduced. We see from eqns. (11), (15), and (16), that neglect of V1 and V2 as well as the off-diagonal Coriolis coupling block-diagonalizes the coupling matrix into four blocks corresponding to the four possible values of ω(±1/2 and ±3/2). For ω = ±1/2, the blocks, of dimension 2, are identical. For ω = ±3/2 the blocks, also identical, are of dimension 1. One consequence is that the two ω = ±3/2 states, which correlate with the 2Π3/2 PES, will not contribute to reaction. This is consistent with the validity of the approximation, used earlier by several groups,[4,33] that cross sections and rate constants for halogen-H2 abstraction reactions determined from calculations on a single PES should be reduced by a factor of 1/2 to account for the flux which, in reality, is associated with the ω = ±3/2 states and hence follows the nonreactive 2Π3/2 PES.
A second consequence is that the dimensionality of the quantum scattering calculation is reduced by a factor of 3, at least in the reactant arrangement, since one need include only two electronic states in the expansion of the wavefunction in eqn. (2), namely the state with ω = ±1/2. Since the difficulty of any close coupling calculation goes as the cube of the number of states (channels) included in the expansion of the wavefunction, this implies that the computational effort will be reduced by at least an order of magnitude. This is not insubstantial.
Because the calculation of the overall reactivity of the excited spin–orbit state of Cl is insensitive to the neglect of Coriolis coupling, calculations of this can be carried out within the computationally faster centrifugal-decoupling (coupled states) approximation.[30] This will further decrease the computational effort, by block-diagonalizing the channel coupling into blocks of one value of K, the body-frame projection of the total angular momentum.
Finally, we see from the discussion in Section II, that nonadiabaticity in the Cl + H2 reaction will be determined by the relative size of the spin–orbit constant (which can be considered a fixed number), and the splitting between the reactive and repulsive, nonreactive, PESs, particularly in the region where these two terms are equal in magnitude. For the ClH2 system, this occurs (Fig. 2) at a Cl–H2 separation of roughly 6 a0, very near the minimum in the van der Waals well for this system. The equilibrium geometry of the Cl–H2 negative ion occurs in linear geometry at a Cl–H2 separation of 5.7 a0.[34] Consequently, the recent photodetachment experiments of Neumark and co-workers[35] will probe the ClH2 PESs in exactly the region responsible for the nonadiabatic transitions which will contribute to the reactivity of Cl in its excited spin–orbit state. As a result, a theoretical simulation of these experiments will provide an invaluable assessment of the accuracy of the CW PESs in this region. Work along these lines is already in progress.