CY) Neary tee tectrons "Problem, Se a) There are several ways to do this, for example TO use “brute force”, or guess the solution and check that its inverse gives the original function. The task is to find the Fourier coefficients Uz so that where # ) for a square lattice ‘Trying to guess the solution, we first note that U(z,y) = —4ip cos (22) cos (222) only involves wavelengths of a. (Recall that k = 2r/A.) Therefore, form, > 1 or ny > 1, Ug = 0. The spatial average of U(2,y) is zero. This is obvious, and can be seen explicitly by calculating the integrals over the period {~a, a] aa | cos (98) de [” eos (222) dy In the reciprocal space, the spatial average of a function is contained in the component K = 0 — think of the Fourier transform which reduces to an integral over coordinate space in that limit. Thus, Uo = 0. To get a product of the form coszcosy, we do not want ne = 0 or my = 0. Additionally, U(z,y) has mirror symmetries. A guess satisfying these requirements is {5% for K = (4%, 42 0, otherwise. We can check this by calculating This is the original function, so the guess is correct. On the other hand, the Fourier coefficients can be obtained fromb) ‘The volume of the primitive cell is now Q = (2x/a)(2m/a) = 4x7/a?. Thus, iff [na] = 1 and |ny| = L. Here, we have used e* = cos + isin to expand the exponentials, and the fact that sinx cos being odd integrates to zero. By the orthogonality of the cosines, the products integrate to zero unless 71 {and n,) is either 1 or —1. One can then utilize cos 2¢ = 2cos?z — 1 to calculate the integrals, or simply multiply the length of the interval 2/a by the average of cos” x which is 3. In trying to find (k;) to first order in U, states that have the same, or similar, free-electron energies are those that contribute significantly. Thus, ‘o(F:) with k, = (£, 2) will couple strongly to components #b(E,) for which ee n?k2 /(Qm) = 7k? /(2m). These are ky =,—1), ky = (-2,4), and ky = (Z,-1). The values of & that must be included in doing pertur- bation theory are, in general, K = (2£n,, ny). Obviously, the three values of #f relating k, to the £,, are (—2£,0), (0,-2£), and (~%£,-2£), From a), we know that Ug = —Up for K = (422,425). For all other values of K, Ug = 0. Of the components (Fy)... (ka), only the component *(f2) with &) = (-2,-%) can be coupled to (Fr): ky =k + (-#,-3). To summarize, the perturbation mixes together free-electron states which are related by K and for which the Fourier coefficient Ug #0. We are interested in what happens to the free-electron state with F = (£, 4). It looks like we need to couple four different states, those with f = (2 4 22,2 + 22), to this state, However, a state has a significant effect (first order in U) if it has thesame, or similar, free-electron energy. Thus, the only important effect on Fy = (7,2) comes from ky = (—%,—), 4 { 3E 7 4) The Schrédinger equation in the subspace involving only (Fx) and (Ra) reads (&, FUR) (£5, — 8) Wea) + Ug, «0h = Notice that &, Fa" e) The eigenvalues of the Hamiltonian are found by solving the characteristic equation, in other words by setting the determinant det|# — ET| to zero. Solving for , we find see 7 i, 0 |= (e. ~¢)?_y2 | Uo & ir (%, 5) m=0 > E=& 4p. ‘The two allowed energies at Bloch index ky are thus € = &%, + Uo and € =&;, ~ Uo. The energy gap is Up. 4) The two lowest bands are sketched below:9). ‘early free electron Fermi surface near a single Bragg plane In the lecture notes, it was shown that when we have two degenerate or nearly degenerate free-electron energy levels, first-order perturbation theory gives for the effect of the weak periodic potential on the energies =} (2+e8,,)= ie ~22,4)] =i To investigate the band structure close to a Bragg plane, it is convenient to measure the wave vector K with respect to the point — 3K on the Bragg plane. Writing 4K + q and resolving 7 into its components parallel (gj) and perpendicular It is also convenient to measure Ep with respect to the lowest value assumed by either of the bands in the Bragg plane, writing Ep = Eo lUgl+A, so that when A <0, no Fermi surface intersects the Bragg plane, a) Let 0< A < 2]lgl, which means that E2, —|Ug| < Er < E%g + \Ugl: In other words, the Fermi energy les in the band gap.\ i \ etaetg | ane | Ee a Ud ‘ | J '\ / intersection -Kk2 k Thus, the Fermi surface lies entirely in the lower band. The intersection of the Fermi surface (lower band) with the Bragg plane is found by setting it equal to € at —1K, in which case gy = 0. This results in &j =O n+ 2 -UgP = daha PDR Fat — Im Only the minus sign qualifies, because A < 2(U/g|. ‘This means that the Fermi surface intersects the Bragg plane in a circle of radius p= \/2mA/n?. b) Im case A > 2|Uig|, the situation changes in the way that we have solutions for both signs. For the lower band, we have again p, = ,/ mAh. For the Fermi surface in the upper band, p = \/2m/R? (A —2IUg|). Thus, the Fermi surface lies in both bands and intersects the Bragg plane in two circles. e \ intersection 4 pnew J intersection ‘The area between the circ : dem, A=n(o} - 3) = ar Mal‘The area of these circles can be measured directly in some metals through the de Haas-van Alpen effect, and |U g| determined directly from experiment for such nearly free electron metals (©) WP Reciprocal lattice a) The primitive vectors for the reciprocal lattice can be chosen to be any set {B1, ba} for which a: 6; = 216,;. The vectors d: = a(1,0) and a = a (4,1) form a centered rectangular lattice. Choosing 5, = (2n/a) (1,-$) and bz = (2n/a)(0, 1) satisfies the requirement. Also the reciprocal lattice is centered. rectangular. ‘The first two Brillouin zones are found by drawing a scetch of the reciprocal lattice, connecting the origin to its nearest, next-nearest, and third-nearest neighbors, and drawing bisectors to these lines. The area, that can be reached from the origin without crossing any of these bisectors (Bragg planes, or lines) forms the first Brillouin zone, the areas that can be reached from the origin by crossing one line form the second Brillouin zone, and so on, b) The first Brillouin zone is the Wigner-Seitz cell in reciprocal space. Because the area of she primitive cell does not depend on the choice of the cell, we may as well calculate it directly fromOn the other hand, it may sometimes be useful to remember that if A’ is the area of the primitive cell in the direct lattice, the primitive cell of the reciprocal lattice has the area (2)?/A'. In this case _ Qn)? _ (Qn)? ~ Ya, x Ge] a? ‘The areas of each Brillouin zones are equal, wm two-dimensional Form surfs Tho primitive vectors of the triangular, or hexagonal, lattice can be taken to be @; a(3,-8) and a = 0 ( 42). Also other choices are possible. ‘The reciprocal lattice is then spun by B, = (2n/a) (1,—Jg) and by = (2n/a) (1, 3g). This is another triangular lattice, rotated by 30 degrees with respect to the direct lattice and with lattice spacing 4m/(V3a) a) The number of &-states in a Brillouin zone is equal to the number of lattice sites and each k-state can contain up to two electrons. If we have six electrons per lattice site, then the electrons fill completely the first three Brillouin zones in k-space. The area of a Brillouin zone is ‘This can be compared with the lattice spacing 4n/(V/3a) ~ 2.309 = b) The reciprocal lattice, the first three Brillouin zones, and (or, rather, disc) are all illustrated below:©) Below is a diagram indicating four sections of the Fermi surface from the Harrison construction. The diagram shows how the surface in the nth Bril- louin zone looks mapped back into the first zone. The image of the interior of the Fermi sphere in the second Brillouin zone is given by all points in the first Brillouin zone that are inside two or more spheres, the image of the interior of the Fermi sphere in the third Brillouin zone by all points in the first Brillouin zone that are inside three or more spheres, etc. The portions of the Fermi surface can then be made continuous by approproate translations through reciprocal lattice vectors. GP Wonnies sanctions ‘The matrix elements of the Hamiltonian between Wannier functions are (RUE) = [aru anu A 1 - far. lz 1 cake = [PELE Po peda Eg ® N Lhe + VV REL Se Sap ror os wl Ege A-B) PERF BAER g / Fo" Wy pA)b) Due to the orthogonality of the Bloch functions Wag ftom different bands, the matrix elements of the Hamiltonian between Wannier functions coming from different bands must vanish. Obviously, the matrix element (F[71|’) depends only upon Ff — &,