Section 15-16 Periodicity in Three Dimensions—Graphite

Section 15-16 Periodicity in Three Dimensions—GraphiteFigure 15-28 Bloch sums based on an s-type AO at each lattice point in the two-dimensional rectangular lattice. (k1, k2) = (a) (0,0), (b) (0, π/a2), (c) (π/a1, π/a2).
situation in Fig. 15-28c. The waves for points X and M can be seen to indeed have the periodicity of the lattice.
Comparison of these three COs leads us to expect lowest energy at , higher energy at X, and highest energy at M. A complete map of the CO energies would require a surface lying over the RFBZ. Instead of showing this surface, it is standard to plot the energy as a function of various straight-line paths connecting special points—e.g., from to X, from X to M, from M to X, from to M. Such line plots are portrayed side by side on a common k axis, as in Fig. 15-29, even though the identity of k (k1, k2, or some combination) changes from panel to panel.
15-16 Periodicity in Three Dimensions—Graphite Graphite is an important commercial material. Some of its uses derive from the fact that it is a very good conductor of electricity—nearly as good as metals. The most stable crystalline form of graphite (Bernal graphite) is depicted in Fig. 15-30a. It comprises

Chapter 15 Molecular Orbital Theory of Periodic Systems
Figure 15-29
Energy versus k along various paths in the RFBZ for the system of s-type AOs in a rectangular two-dimensional lattice.
carbon atoms in two-dimensional sheets having hexagonal symmetry. These are stacked so that half of the carbons in a sheet lie directly between carbons in adjacent sheets while the other half lie between empty hexagon centers. The atoms within a sheet are strongly covalently bonded to their neighbors, with nearest-neighbor bond lengths of 1.42 Å. The sheets are separated by 3.5 Å, indicative that the forces between them are very weak. (This explains why they fracture into sheetlike fragments that can slide over each other like a pile of playing cards, making ground graphite a good lubricant.)
A sensible approach to understanding the electronic structure of graphite, then, is to first analyze the two-dimensional sheet and then consider the perturbation involved in forming layers of sheets.
A section of a two-dimensional sheet, the translation vectors a1 and a2, and the primitive unit cell are sketched in Fig. 15-30b. The primitive unit cell contains two carbon atoms. The angle between a1 and a2 is 120◦. Notice that the lattice points in real space, at the vector origin and at the termini of integral numbers of steps from the origin, fall at the centers of hexagons, where there are no atoms. This illustrates that there is no particular identification of direct lattice points with atoms or other structural features of the molecules that constitute a crystal. In the present instance the direct lattice, shown in Fig. 15-30c, looks like a graphite crystal that has been rotated by 30◦ with an extra point in the center of each hexagon. In general, even though the direct lattice need not look identical to the crystal, it will have the same symmetry as the crystal, in this case hexagonal. In Fig. 15-30d is shown the relation between the vectors a and the reciprocal space vectors b. Since b1 must be perpendicular to a2, etc., it follows that the angle between b vectors is 60◦. The reciprocal lattice generated by the b vectors appears in Fig. 15-30e and is again a centered-hexagon pattern, but rotated with respect to the direct lattice. The FBZ is constructed by drawing lines from one reciprocal lattice point to all its nearest neighbors and then cutting them with perpendicular bisectors. This yields a hexagon, as shown in Fig. 15-30f. Also shown is the RFBZ. The entire FBZ can be filled by putting the RFBZ through the symmetry operations of the hexagonal sheet, so it includes all the points not equivalent by symmetry.
Points of special interest in the RFBZ are labeled , M, and K. The point corre sponds to the (k1, k2) values (0,0). This is a unique point in the FBZ, so it produces one nondegenerate Bloch sum for each basis function in the unit cell. The point M


Section 15-16 Periodicity in Three Dimensions—Graphite
a1
a2 (a)
(b)
a1
b1
b1
a
b
2
b
2
2
(c)
(d)
(e)
Figure 15-30 (a) The crystal structure of Bernal graphite. Each point represents a carbon atom.
(b) Top view of a hexagonal sheet of graphite showing the unit cell and translation vectors for the two-dimensional structure. The “origin point” of the vectors does not correspond to an atom location.
(c) The direct lattice produced by translations of the “origin point” of the a vectors. The full lines emphasize that this is a lattice of centered hexagons. The graphite structure is indicated with dashed lines. (d) Reciprocal lattice vectors b are shown in relation to vectors a. (e) The reciprocal lattice, produced by translating the origin point of the b vectors. (f) Construction of the FBZ by bisecting lines connecting one reciprocal lattice point to all of its nearest neighbors. The RFBZ is shaded.
(g) Geometry of the relation between point K and the vectors connecting it to the origin.
corresponds to (π/a, 0). We might expect this to be a six-fold degenerate point because it is one of six symmetrically equivalent points in the FBZ. However, the point diametrically opposite this in the FBZ, at (−π/a, 0), does not produce a separate independent Bloch sum because, if we combine these complex functions to form a real (cosine) Chapter 15 Molecular Orbital Theory of Periodic Systems wave, the other real (sine) wave puts a node at every real lattice point, meaning that the basis function is multiplied everywhere by zero, which is not acceptable. This is similar to what we saw in the one-dimensional case, which led us to exclude −π/a from the range of k. So here again we recognize that two points on opposite sides of the FBZ where perpendicular bisectors intersect lines between reciprocal lattice points really only account for one independent state. This means that the point M represents three independent Bloch sums rather than six. We shall have more to say about this later. The point labeled K does not lie on a b vector, so must be expressible as a sum of two vectors. Figure 15-30g shows the relevant construction. Since the lowest ray has length π/a, the central ray has length π/(a cos 30◦). Then the length of k1 or k2 must be π/a(2 cos2 30◦), or 2π/3a. So (k1, k2) = (2π/3a, 2π/3a) at K. As will be explained later, K does not have reduced degeneracy as M does.
Since there are two carbon atoms per unit cell and four valence AOs per carbon, we have eight basis functions per unit cell for a minimal valence basis set calculation.
Therefore, we expect eight bands. Since the system is planar we can distinguish two π AOs and six σ AOs. We know that these two symmetry types will not interact for any k, so we can predict or interpret the σ and π band behaviors independently.
Let us begin by making some predictions about the two π bands. As was the case for polyacetylene, the two basis sets that are simplest to work with are the bonding (πu) and antibonding (πg) combinations in a unit cell. We shall label the band resulting from the bonding function π and the other band π ∗. At the point , the unit cell basis functions are simply translated along a1 and a2 with no change in phase, yielding the COs sketched in Fig. 15-31a. The π CO is bonding in every bond, π ∗ is antibonding, so these will be widely separated in energy. Indeed, there is no way they could become more widely separated by intermixing, so we can assume that, if we were to do a variational calculation, these Bloch sums would not be modified.
At point M, translation by each step along a1 is accompanied by phase reversal because k1 = π/a1, whereas translation along a2 involves no change because k2 = 0 (Fig. 15-31b). Now we find each carbon to be bonding to two neighbors and antibonding to the third in the π CO and the reverse in the π ∗CO. Once again, it is not hard to see that any mixing of these two would decrease their energy difference (Problem 15-24), so these Bloch sums should be unchanged in a variational calculation. Therefore, the π and π ∗ bands should be separated in energy at M, but only by about one third as much as at . As we mentioned above, the COs sketched in Fig. 15-31b are each degenerate with two other COs corresponding to symmetrically equivalent points in the FBZ, for example, the points (0, π/a) and (π/a, −π/a). Sketching these (Problem 15-25) shows that they are like the CO already sketched except rotated by ±60◦.
Point K is more complicated to analyze because it has k values that do not produce COs having the periodicity of the crystal. This means that the decreased degeneracy we find for high-symmetry points like X in Section 15-15, caused because one of the real coefficient waves has a node at every unit cell, does not occur at K. (If the wave lacks the periodicity of the crystal, it cannot put a node at every unit cell, by definition.)
That means we will have two π and two π ∗ solutions at K. Also, the fact that these waves lack the periodicity of the crystal makes them more complicated to sketch.
Nevertheless, undaunted, we plunge ahead. Since K corresponds to (2π/3a, 2π/3a), we expect steps along either a1 or a2 in real COs to be modulated by either cos(2nπ/3) or sin(2nπ/3). For n = 0, 1, 2, 3, this gives the following repeating series for the cosine

Section 15-16 Periodicity in Three Dimensions—GraphiteFigure 15-31 π and π ∗ COs at the point (a) , (b) M, (c) K.
and sine cases, respectively: 1, −0.5, −0.5, 1 and 0, 0.866, −0.866, 0. The results of using these factors to modulate translations of the πu and πg basis functions are sketched in Fig. 15-31c. Inspection of these sketches shows that the extent of bonding and antibonding interactions around any carbon exactly cancels in every case. This means that the π and π ∗ bands are degenerate at K. Each of the pairs of COs sketched in Fig. 15-31c is degenerate with two other pairs that are rotated by ±60◦. Therefore, the degeneracy of this energy level is 12.
The band diagram for two-dimensional graphite is sketched in Fig. 15-32 for a circuit around the RFBZ. The π and π ∗ bands behave essentially as we anticipated, being widely separated at , much less separated at M, and degenerate at K. We note that there are three σ bands at lower energy and three more at higher energy, corresponding to predominantly bonding and antibonding situations, respectively. However, these bands are less easily analyzed because the 2s, 2px, and 2py AOs undergo varying extents of mixing with each other as k varies. That is, these are bands where the variational method causes extensive mixing of Bloch sums. Nevertheless, we can see patterns that would not be difficult to investigate. For example, we can see that the π band and one of the σ bands run up from to M, whereas the other two σ bands run down. This is consistent with the fact that the 2s and 2pπ AOs relate to corresponding AOs in adjacent unit cells in one way, while 2pσ AOs relate in the opposite way (Problem 15-28).

Chapter 15 Molecular Orbital Theory of Periodic SystemsFigure 15-32 Valence-band diagram for two-dimensional graphite. (Adapted from Painter and Ellis [6].)
There are eight valence electrons per unit cell in neutral graphite. For the band diagram of Fig. 15-32 it is simple to figure out the band-population scheme. Eight electrons require four bands, and we see that the four lowest-energy bands lie below the other four bands for all values of k. This means that the four lowest-energy bands are filled and the others are empty. Therefore, the Fermi level comes at the top of the π band, which occurs at K. We see that there is an empty band (π ∗) at the same energy, so there is no gap at the Fermi level. Thus the two-dimensional lattice is predicted to be a good electrical conductor. We could examine the possibility that lattice deformation would split the levels to produce a gap (a Peierls distortion), but, since we have structural information indicating that graphite does in fact have hexagonal symmetry, we will not pursue that point.
If there were crossing of one of the σ ∗ COs with the π CO over part of the range of k, then some fraction of the electronic population would occupy the low-energy portion of the σ ∗ band at the expense of the high-energy portion of the π band. Analysis of this would require more effort, but the result would still predict good conductivity since, instead of a full band “touching” the bottom of an empty band, we would have two partially filled bands.
Next we consider the effects of stacking two-dimensional sheets to form the three dimensional crystal.4 As Fig. 15-30a indicates, the stacking pattern repeats the orientation of a sheet after one intervening layer in an ABABAB stacking pattern. This means that the three-dimensional unit cell must contain carbon atoms from two layers.
The unit cell now has three associated translation vectors. The two “intrasheet” translations a1 and a2 are as before, and the “intersheet” translation a3 is perpendicular to the planes of the sheets and 7.0 Å long. Because a3 is the longest vector in real space, b3 is the shortest vector in reciprocal space, leading to a reciprocal lattice where sheets 4A discussion of effects due to other stacking arrangements can be found in LaFemina and Lowe [7].

Section 15-16 Periodicity in Three Dimensions—Graphite of centered hexagons are layered with short intersheet distances. The resulting FBZ, produced by perpendicular bisecting planes (since we are now working in three dimensions), appears in Fig. 15-33a. The RFBZ is similar to that in Fig. 15-30f except that it now has a third dimension, over which k3 ranges from 0 to π/a3 (Problem 15-26).
b3
b1
π
π
L , 0,
π
a1
a3
A 0, 0, π
a3
M , 0, 0
a1
2π 2π π
H
,
,
3a
RFBZ
1 3a2 a3 2π 2π
K
,
, 0
3a1 3a2 Γ(0, 0, 0)
b2 (a)
a3
(b)
Figure 15-33 (a) The FBZ and RFBZ for three-dimensional Bernal graphite. (b) View of the unit cell for Bernal graphite, as seen along an axis parallel to the hexagonal sheets. (Compare to Fig. 15-30a.)
Chapter 15 Molecular Orbital Theory of Periodic Systems
Let us see if we can predict the modifications in energies of COs that should occur at some of the special points in the RFBZ as a result of intersheet interactions. Because three-dimensional Bernal graphite is symmetric for reflection through the plane of a hexagonal sheet, we continue to have a valid distinction between σ and π COs. We expect intersheet interactions to be small, so we can take a perturbational approach.
The interactions involving π COs should be largest since pπ AOs project farthest from the plane, so we will restrict our attention to the π bands.
A side view of the unit cell appears in Fig. 15-33b. Obviously, some of the intersheet interactions are already included in the unit cell, and the others will occur between unit cells. We will first consider the energy changes due to new interactions within the cell.
These do not depend on k values. Then we will consider energy changes caused by stacking the unit cells. These depend on k3. We will ignore all interactions except those between carbons directly above or below each other in adjacent sheets.
We begin with the situation where the π and π ∗ bands are nondegenerate, e.g., at or M. We know from our two-dimensional treatment that each covalently bonded pair of carbons in the unit cell will have a πu MO when we are dealing with the π band and a πg MO for the π∗ band. (See Fig. 15-31a, b.) But we have a choice of arranging these in the unit cell to give long-range bonding or antibonding between layers (Fig. 15-34).
These two arrangements cause the energies to rise or fall slightly (indicated by E ↑ or E ↓), splitting both the π and π∗ bands.
Next we examine what happens when we stack these unit cells. Consider first the point . Here k3 = 0, so we stack the unit cells with no change. If we start with πanti of Fig. 15-34a, we obtain Fig. 15-35a, which shows that the unit cells interact in an antibonding manner. Here, then, is a Bloch sum that is antibonding within and also between unit cells. The symbol E ↑↑ indicates two destabilizing interactions. Each arrow represents the same amount of energy because the interlayer distance within a unit cell is the same as that between unit cells and also because there are equal numbers of inter- and intralayer interactions in the crystal. Similar sketches easily demonstrate how the other three unit-cell function sets of Fig. 15-34 behave at : πbond, E ↓↓; π∗ , E ↑↑; π∗ , E ↓↓. Therefore, at , the π and π∗ bands are both
anti
bond
split.
Now let us consider point A. Here we stack the unit cells together with a reversal in sign. The situation for πanti is shown in Fig. 15-35b. The intercell interaction has reversed from the k3 = 0 case. This means that the E ↑↑ and E ↓↓ cases from now are E ↑↓ and E ↓↑. The levels are not split by interlayer effects at A.
When we consider points M and L, we again are dealing with the same πu and πg functions in the unit cell (see Fig. 15-31b), so we obtain the same results with k3 = 0, or π/a3. The π and π∗ bands are split at M by the same amount as at , and at L they are not split at all.
At point K, the π and π ∗ bands are degenerate. This means we must ask which are the proper zeroth-order two-dimensional COs with which to analyze the perturbation.
We know that these will be four COs produced by mixing the four COs shown in Fig. 15-31c. This mixing is determined by solving a 4 × 4 determinantal equation, which, as was pointed out in Chapter 12, is the same equation we would use to discover the COs having minimum and maximum energy changes as a result of the perturbation.
As a shortcut, then, we can simply look for the COs that respond most differently. Our analysis at and M tells us that the πu and πg bases will give a splitting of all the

Section 15-16 Periodicity in Three Dimensions—Graphite
Figure 15-34 The four independent arrangements of π (or πu) and π ∗ (or πg) functions in a unit cell. Labels “anti” and “bond” refer to the interactions between AOs directly above/below each other.
E ↑ and E ↓ indicate whether these interactions raise or lower the energy.
COs. Not bad, but we can do even better by mixing the COs pictured in Fig. 15-31c.
Adding and subtracting the cos π ∗ and cos π COs produces two new functions whose unit cells (in two dimensions) have a 2pπ AO on one carbon and nothing on the other, and vice versa. (In effect, we are taking the sum and difference of the πu and πg basis functions that we constructed in the first place.) These new basis functions give us four new unit cell bases for the three-dimensional case, shown in Fig. 15-36. Two of these place all the 2pπ AOs on atoms that lie directly above and below each other, so these rise or fall in energy due to the intracell interactions. The other two place all the 2pπ AOs on carbons that do not have neighbors directly above or below them, so these do not undergo energy change. The rise and fall here are larger than for the cases pictured in Fig. 15-34 because the AO coefficients are larger when the basis set exists on

Chapter 15 Molecular Orbital Theory of Periodic Systems
Figure 15-35 (a) The result of translating a πanti unit-cell basis function one step along a3 with no change of sign or value (k3 = 0). (b) Same as (a) except now sign reverses (k3 = π/a).
fewer atoms. These, then, are the proper zeroth-order functions with which to analyze the stacking perturbation. (The argument here is equivalent to saying that a calculus min-max problem will select roots −2, 0, 0, +2 in preference to −1, −1, +1, +1.)
When we put these unit-cell bases together at K, where k3 = 0, we find that the func tions that do not interact within the cell do not interact between cells either (Fig. 15-37a).
These two cases remain unsplit in energy. The cases that interact within the cell interact again and in the same way between cells, so these two cases are split even more (Fig. 15-37b). At L, the intercell interactions reverse but the intracell interactions are unchanged. The E· cases are unaffected, but E ↑ and E ↓ become E ↑↓ and E ↓↑ (Fig. 15-37c), just as at A and H , so the splitting disappears at L.
A sketch of the π and π ∗ bands for a route around the edges of the RFBZ appears in Fig. 15-38. In general, no splitting of these bands occurs for points on the top of the RFBZ. For points on the bottom, splitting is larger at K than at or M.
Computing the electronic energy per unit cell for graphite requires knowing the average energy of the occupied bands, not only over the paths shown but also over the interior points of the RFBZ. The method of Chadi and Cohen [5] can be used to select

Section 15-16 Periodicity in Three Dimensions—GraphiteFigure 15-36 The four independent unit-cell functions that result from requiring maximum difference in response to the interlayer interaction. E· indicates no energy change.
an optimized set of weighted points in the RFBZ for purposes of calculating accurate average energies.
One might wonder whether the splittings depicted in Fig. 15-38 have any detectable consequences. One change that occurs as a result of this splitting is that the Fermi level, EF, now refers to π electrons in COs that are not split along K–H by the perturbation, i.e., in COs like the one pictured in Fig. 15-37a. This means that the electrons having the highest energy are predicted to be in 2pπ AOs at carbons that do not have carbon atoms directly above or below them in the crystal. This comes about because the electrons that do have such neighbors are lowered slightly in energy due to weak bonding between layers. This is relevant because there exists an experimental technique called scanning tunneling microscopy (STM) that detects the locations of surface electronic charge having energy near the Fermi level. When STM maps are made of electron distribution at a clean graphite surface, a trigonal pattern is seen, rather than the expected hexagonal one. Evidently, not all the atoms are “seen” equally well. The trigonal pattern is