15-14 Energy Calculations
Chapter 15 Molecular Orbital Theory of Periodic Systems15-14 Energy Calculations
Suppose we wish to compare the calculated energy of regular polyacetylene to that of a bond-alternating version, or that of PPP with PPP-N2. How do we get energies from the band calculations for these two cases? In a H¨uckel-type calculation of a molecule, we simply add up the one-electron energies. In a band calculation we are faced with a very large set of one-electron energies—one for each k value. How do we deal with this?
Consider again the simple H¨uckel band diagram for regular polyacetylene (Fig. 15-17a). This band diagram indicates that electrons in the π CO at k = 0 have an energy of α + 2β, those at k = π/a have E = α, and those at k = π/2a have an energy somewhat below a + β. It is fairly obvious that the CO average energy is somewhat below α + β. Each CO has two electrons delocalized over the polymer, but the net number of electrons per unit cell is two. Therefore, with an average π CO energy of below a + β and two π electrons per unit cell, we have a π energy per unit cell below 2a + 2β. To be more precise would require a more precise average energy for the CO.
[Even without such a calculation, however, we can see that bond alternation causes the average energy of the occupied π band to drop (Fig. 15-17b), so the π -electron energy per unit cell is lower for the alternating structure.]
The problem, then, is to calculate an accurate average energy for each of the filled bands, the average being over the first Brillouin zone, and then multiply each such average energy by the number of electrons in that band, per unit cell. The sum of these is the H¨uckel total energy per unit cell. (This procedure assumes that the bands are not partially filled.)
The practical difficulty in doing this is that each variational calculation is carried out at a single point in k space. Thus, for the variational bands sketched in Fig. 15-23, an EHMO calculation was made at k = 0, another was made at k = π/2a, and a third was made at k = π/a. Then lines were drawn to connect the energy points found at these k values, with consideration being given to the noncrossing rule. Obviously, this produces a rather qualitative diagram. To refine it would require making additional variational calculations at intermediate points in k space. Thus, we must be concerned with the trade-off between accuracy of final average energy and effort needed to achieve well-characterized band energies from which to calculate that average.
Achieving accurate average CO energies from values at a few k points is possible if those points are sensibly chosen and if appropriate weighting factors are employed. The problem of choosing the correct few points and their weight factors has been worked out by Chadi and Cohen [5] for multidimensional systems of various symmetries. Although this is a matter of real practical interest, we will not explore it further here.
15-15 Two-Dimensional Periodicity and Vectorsin Reciprocal Space
In one-dimensional problems there is a single translation direction, a single step size a, and a single accompanying variable k. It is a simple matter in such problems to treat a and k as scalars and to plot band energies versus k in the range of the first Brillouin zone, −π/a