It is rather obvious that one-dimensional theory is not good enough to comprehend quantum mechanical tunneling phenomena. For instance, the tunneling splitting estimated using the one-dimensional potential along the intrinsic reaction coordinate gives a few orders of magnitude smaller value [56]. There is a more crucial example that demonstrates the conceptual deficiency of one-dimensional treatment and shows interesting features peculiar to multidimensional tunneling phenomena. This is explained in Chapter 4. In this sense, multidimensional treatment is definitely needed. Herewe describe the two-dimensionalWKB theory developed by Takada and Nakamura [30].
It is rather obvious that one-dimensional theory is not good enough to comprehend quantum mechanical tunneling phenomena. For instance, the tunneling splitting estimated using the one-dimensional potential along the intrinsic reaction coordinate gives a few orders of magnitude smaller value [56]. There is a more crucial example that demonstrates the conceptual deficiency of one-dimensional treatment and shows interesting features peculiar to multidimensional tunneling phenomena. This is explained in Chapter 4. In this sense, multidimensional treatment is definitely needed. Herewe describe the two-dimensionalWKB theory developed by Takada and Nakamura [30].
Tunneling is a supreme quantum effect. Every introductory text [1] on quantum mechanics gives the paradigm example of a particle tunneling through a one-dimensional potential barrier despite having a total energy less than the barrier height. Indeed, the reader typically works through a number of exercises, all involving one-dimensional potential barriers of one form or another modeling several key physical phenomena ranging from atom transfer reactions to the decay of α-particles [2]. However, one seldom encounters coupled multidimenisonal tunneling in such texts since an analytical solution of the Schro¨dinger equation in such cases is not possible. Interestingly, the richness and complexity of the tunneling phenomenon manifest themselves in full glory in the case of multidimensional systems [3]. Thus, for instance, the usual one-dimensional expectation of increasing tunneling splittings as one approaches the barrier top from below is not necessarily true as soon as one couples another bound degree of freedom to the tunneling coordinate. In the context of molecular reaction dynamics, multidimensional tunneling can result in strong mode-specificity and fluctuations in the reaction rates [4]. In fact, a proper description of tunneling of electrons and hydrogen atoms is absolutely essential [5,6] even in molecular systems as large as enzymes and proteins. Although one usually assumes tunneling effects to be significant in molecules involving light atom transfers, it is worth pointing out that neglecting the tunneling of even a heavy atom like carbon is the difference between a reaction occurring or not occurring. In particular, one can underestimate rates by nearly hundred orders of magnitude [7]. Interestingly, and perhaps paradoxically, several penetrating insights into the nature and mechanism of multidimensional barrier tunneling have been obtained from a phase-space perspective [8,9]. The contributions by Creagh, Shudo and Ikeda, and Takahashi in the present volume provide a detailed account of the latest advances in the phasespace-based understanding of multidimensional barrier tunneling.
As was pointed out in Section 3.1, there are some interesting features in multidimensional tunneling compared with the one-dimensional case [30,31,46]. These are the following: (1) There exist two types of tunneling, pure tunneling and mixed tunneling. (2) No tunneling path can be defined in the case of mixed tunneling. (3) Vibrational excitation can suppress tunneling because of mixed tunneling. In this section, these peculiar features are numerically demonstrated and analyzed [31] by taking tunneling splitting in symmetric double well potential as an example. Three typical two-dimensional model potentials are used.
In the case of tunneling in chemical reactions [see figure1.1(c)], the total energy of reagents cannot be generally considered to be small; treating it as the first-order term of h¯ as was successfully done in Chapters 6 and 7 does not lead to any sensible semiclassical theory. The semiclassical mechanics requires (1) construction of Lagrange manifolds, (2) determination of amultidimensional caustic surface, and (3) connection of wave functions between classically allowed and forbidden regions [59]. Unfortunately, these problems have not yet been solved mathematically. As was discussed in Section 3.1, in the framework of the ordinary WKB approach we can construct Lagrange manifolds and caustics in the case of two dimensions and the connection problem can be treated relatively well by assuming the local separability. Extension of this method to higher dimensions, however, cannot be practical. The complexvalued (complex coordinate and/or momentum) WKB theory has been devised, as briefly explained there, but cannot be simply extended to multidimensions (see, for instance, [15]). This formidable task can hardly be accomplished mainly because of the complexity of the Lagrange manifold and the fact that analytical properties of any realistic potential energy surfaces extended into the complex-valued coordinate space cannot be guaranteed. The use of a complex trajectory was nicely materialized by the double-ended complex trajectory in the classical S-matrix theory [206], in which the quantized boundary conditions are satisfied at both (initial and final) ends. Unfortunately, this method turned out not to be practically useful, because finding such trajectories is also a formidable task in a multidimensional complex coordinate (and/or momentum) space. We can, however, make a break through the difficulty of detecting caustics in multidimensional space. An efficient way of doing that has been devised along classical trajectory [47]. This is not a trivial task at all, since it is not possible to find caustics in high-dimensional space by just shooting classical trajectories. Since the caustics can be found along each classical trajectory, chemical dynamics with tunneling effects incorporated can now be treated by the on-the-fly method. In this subsection, this method is explained.
Abstract One-dimensional (1D) systems provide examples of many of the most import- ant features of quantum mechanics, but it is also instructive to consider two-dimensional (hereafter 2D or planar) systems for several reasons:Systems with two spatial degrees of freedom provide more opportunities to study multivariable probability concepts, separation of coordinates tech- niques, and new mathematical methods and special functions. They also allow for the visualization of many quantum phenomena which arise in more real- istic three-dimensional (3D) systems, but which are obviously difficult to plot in 3D.
There are enormous amounts of literature dealing with quantum mechanical tunneling phenomena, andmanymonographs have also been published recently covering various fields of the phenomenon [21-24]. Unfortunately, however, theories of multidimensional tunneling phenomena are not yet satisfactorily developed. Effective numerical algorithms in quantum mechanics such as MCTDH (multichannel timedependent Hartree) method [25] and quantum Monte Carlo method [26] have been developed and successfully implemented to study quantum dynamics in multidimensional systems. However, exact quantum dynamics methods require a formidable task including accurate ab initio calculations of global potential energy surfaces. Besides, it is not straightforward to obtain the physical insights of dynamic processes.
One of the appealing features of tunneling as a theoretical problem is that it poses interesting questions regarding the complexified dynamics of the classical limit. Often, these questions are of a different nature to the ones that arise in a purely dynamical systems context and can therefore push the dynamics side of the problem in novel directions, without necessarily being technically hard. In this article we explore a number of scenarios where these underlying problems of complex dynamics can be well enough understood to allow explicit analytical approximations to be given for tunneling rates. The treatment here is not intended to be comprehensive and short shrift is given to very interesting and important tunneling regimes, such as those of resonance-and chaos-assisted tunneling [3,14,16,17,40,48,78], for example. Instead, by concentrating on relatively tractable problems where answers can be given in terms of simple geometrical characterizations of the complex dynamics, we aim to achieve an understanding of problems which are important in their own right but which may also later serve as building blocks in the solution of more difficult problems.