Skip to article frontmatterSkip to article content
Site not loading correctly?

This may be due to an incorrect BASE_URL configuration. See the MyST Documentation for reference.

Homework 3 — From Local Symmetry to Effective Hamiltonians

National Tsing Hua University

Learning objectives

In this homework, we will approach effective Hamiltonians from two complementary directions:

Orbitals + geometry + symmetry⟶H⟶energy bands,\text{Orbitals + geometry + symmetry}\longrightarrow H\longrightarrow \text{energy bands},

and

energy bands⟶?H.\text{energy bands}\overset{?}{\longrightarrow}H.

After completing the homework, you should be able to:

  1. construct real dd orbitals from l=2l=2 spherical harmonics and use crystal symmetry to constrain local Hamiltonians;

  2. derive symmetry-based selection rules for hopping matrix elements;

  3. identify orbital combinations that cannot hybridize;

  4. reconstruct simple tight-binding parameters from band energies;

  5. explain why energy bands alone do not generally determine a unique Hamiltonian.

Instructions. Show your reasoning, not just your final answers. You may use a short numerical program to check your results, but analytical explanations are required. Unless stated otherwise, neglect spin and spin–orbit coupling. Energies are in eV where indicated.


Problem 1 — Ligand-field theory: symmetry before diagonalization

Consider a transition-metal ion at the origin surrounded by six identical ligands at

(±a,0,0),(0,±a,0),(0,0,±a).(\pm a,0,0),\qquad (0,\pm a,0),\qquad (0,0,\pm a).

For an isolated atom with a spherically symmetric potential, a single-electron wavefunction can be written as

ψnlm(r)=Rnl(r)Ylm(θ,ϕ).\psi_{nlm}(\mathbf r)=R_{nl}(r)Y_l^m(\theta,\phi).

We will first recover the familiar real dd orbitals from the l=2l=2 spherical harmonics, then use the symmetry of the ligand environment to constrain the Hamiltonian. Throughout, use

x=rsin⁡θcos⁡ϕ,y=rsin⁡θsin⁡ϕ,z=rcos⁡θ.x=r\sin\theta\cos\phi,\qquad y=r\sin\theta\sin\phi,\qquad z=r\cos\theta.

(a) From spherical harmonics to crystal-field splitting

The five normalized l=2l=2 spherical harmonics (Condon–Shortley convention) are

Y20=516π(3cos⁡2θ−1),Y_2^0=\sqrt{\frac{5}{16\pi}}(3\cos^2\theta-1),
Y2±1=∓158πsin⁡θcos⁡θ e±iϕ,Y_2^{\pm1}=\mp\sqrt{\frac{15}{8\pi}}\sin\theta\cos\theta\,e^{\pm i\phi},
Y2±2=1532πsin⁡2θ e±2iϕ.Y_2^{\pm2}=\sqrt{\frac{15}{32\pi}}\sin^2\theta\,e^{\pm2i\phi}.
  1. Atomic degeneracy. Why are all five l=2l=2 states degenerate for an isolated atom with a central potential (neglecting spin–orbit coupling)? What continuous symmetry is present? Why is it reduced by the six ligands?

  2. Constructing real dd orbitals. Express the five Y2mY_2^m in terms of Cartesian coordinates (up to normalization). Then construct five real, normalized linear combinations of them whose angular dependence is proportional to

    xyr2,yzr2,zxr2,x2−y2r2,3z2−r2r2.\frac{xy}{r^2},\quad\frac{yz}{r^2},\quad\frac{zx}{r^2},\quad \frac{x^2-y^2}{r^2},\quad\frac{3z^2-r^2}{r^2}.

    Denote the corresponding orbitals by dxyd_{xy}, dyzd_{yz}, dzxd_{zx}, dx2−y2d_{x^2-y^2}, and d3z2−r2d_{3z^2-r^2}. You may ignore the common radial factor. (Hint: Form sums and differences of Y2mY_2^m and Y2−mY_2^{-m}; suitable factors of ii produce real functions.)

  3. Why change basis? Consider the rotation

    C4z:(x,y,z)↦(−y,x,z).C_{4z}:(x,y,z)\mapsto(-y,x,z).

    Determine how the five real dd-orbital polynomials transform. Compare with the transformation of Y2mY_2^m under a rotation about zz. Why might the real-orbital basis be more convenient for discussing ligands placed along Cartesian axes? (You need not construct the full 5×55\times5 rotation matrix.)

  4. Finding invariant subspaces. Now also consider

    C3:(x,y,z)↦(y,z,x).C_3:(x,y,z)\mapsto(y,z,x).

    Determine the transformation of the five real orbital polynomials under C3C_3. Using both rotations, show that the two subspaces

    Ht=span⁡{dxy,dyz,dzx},He=span⁡{dx2−y2,d3z2−r2}\mathcal H_t=\operatorname{span}\{d_{xy},d_{yz},d_{zx}\},\qquad \mathcal H_e=\operatorname{span}\{d_{x^2-y^2},d_{3z^2-r^2}\}

    do not mix. (Hint: The two orbitals in He\mathcal H_e can transform into linear combinations of one another.)

  5. Constrained local Hamiltonian. Under the full octahedral symmetry group OhO_h, these are inequivalent irreducible subspaces, conventionally labeled t2gt_{2g} and ege_g. Explain why a symmetry-preserving, spin-independent local Hamiltonian has the form

    HCF=ϵtI3⊕ϵeI2.H_{\mathrm{CF}}=\epsilon_t I_3\oplus\epsilon_e I_2.

    How many independent on-site energies remain? Does symmetry alone tell you whether ϵe>ϵt\epsilon_e>\epsilon_t? (You may use the stated irreducibility of the two subspaces; a full group-character calculation is not required.)

  6. Physical interpretation. In the usual electrostatic picture of negatively charged ligands along the coordinate axes, which set of orbitals experiences stronger repulsion? Explain from their spatial orientation, without computing an electrostatic integral. Why is this conclusion additional physical input rather than a consequence of symmetry alone?

Takeaway: The same five-dimensional atomic orbital space can be described in different bases. A basis adapted to the crystal symmetry reveals the block structure of the local Hamiltonian.

(b) Which electrons are allowed to hop?

Now consider two metal sites AA and BB joined by a bond along the xx direction. Retain two orbitals on each site, dxyd_{xy} and dxzd_{xz}, and assume that the bond environment is invariant under both reflections My:y↦−yM_y:y\mapsto-y and Mz:z↦−zM_z:z\mapsto-z.

  1. Orbital parity. Determine the eigenvalue (+1 or -1) of each orbital under MyM_y and MzM_z.

  2. Selection rules. Write the hopping Hamiltonian as

HAB=∑α,β∈{xy,xz}(tαβcAα†cBβ+h.c.).H_{AB}=\sum_{\alpha,\beta\in\{xy,xz\}} \left(t_{\alpha\beta}c^\dagger_{A\alpha}c_{B\beta}+\mathrm{h.c.}\right).

Which elements of the 2×22\times2 hopping matrix T=(tαβ)T=(t_{\alpha\beta}) must vanish by symmetry? Explain your answer without evaluating any overlap integrals. Are the two diagonal hopping amplitudes required to be equal?

Takeaway: A forbidden hopping amplitude can be exactly zero because of symmetry, not because the orbitals are far apart.

(c) An orbital that refuses to hybridize

Consider a simplified three-orbital metal–ligand cluster with one metal orbital ∣d⟩|d\rangle and two ligand orbitals ∣p1⟩,∣p2⟩|p_1\rangle,|p_2\rangle. In the ordered basis (∣d⟩,∣p1⟩,∣p2⟩)(|d\rangle,|p_1\rangle,|p_2\rangle), let

H=(ϵdt−ttϵp0−t0ϵp),H=\begin{pmatrix} \epsilon_d&t&-t\\ t&\epsilon_p&0\\ -t&0&\epsilon_p \end{pmatrix},

where tt is real. Suppose the cluster has a symmetry SS that acts as S∣d⟩=−∣d⟩S|d\rangle=-|d\rangle, S∣p1⟩=∣p2⟩S|p_1\rangle=|p_2\rangle, and S∣p2⟩=∣p1⟩S|p_2\rangle=|p_1\rangle.

  1. Define the ligand combinations

∣p±⟩=∣p1⟩±∣p2⟩2.|p_\pm\rangle=\frac{|p_1\rangle\pm|p_2\rangle}{\sqrt2}.

Rewrite HH in the basis (∣d⟩,∣p+⟩,∣p−⟩)(|d\rangle,|p_+\rangle,|p_-\rangle). Which ligand combination is an exact eigenstate for arbitrary tt? What is its energy?

  1. Determine the symmetry eigenvalues of ∣d⟩|d\rangle, ∣p+⟩|p_+\rangle, and ∣p−⟩|p_-\rangle under SS. Explain the absence of hybridization in terms of symmetry sectors.

  2. Challenge: robustness. Change the ligand energies to ϵp+δ\epsilon_p+\delta and ϵp−δ\epsilon_p-\delta, leaving the hoppings unchanged. Does the previously decoupled ligand combination remain an exact eigenstate when δ≠0\delta\ne0? Identify the symmetry that has been broken and explain your conclusion.

Inspiring question: Can a state remain localized even when nonzero hopping pathways are present? What is the difference between weak hopping and symmetry-forbidden hybridization?


Problem 2 — From band energies to tight-binding Hamiltonians

Imagine that an electronic-structure calculation has produced the energy bands of a one-dimensional crystal. Your task is to infer a simple effective tight-binding description. Throughout this problem, aa is the lattice constant and kk lies in the Brillouin zone [−π/a,π/a][-\pi/a,\pi/a].

(a) Reconstruct a one-band model

The calculated energies for an isolated band are:

kaka0π/3\pi/3π/2\pi/22π/32\pi/3π\pi
E(k)E(k) (eV)-2.0-1.00.01.02.0

Assume one orbital per unit cell and the real-space Hamiltonian

H=ϵ0∑jcj†cj−t1∑j(cj+1†cj+h.c.).H=\epsilon_0\sum_j c_j^\dagger c_j -t_1\sum_j(c_{j+1}^\dagger c_j+\mathrm{h.c.}).
  1. Use a Fourier transform to derive E(k)E(k).

  2. Determine ϵ0\epsilon_0 and t1t_1 from the data. Check your answer against all five data points.

  3. Now include second-neighbor hopping,

    H2=−t2∑j(cj+2†cj+h.c.).H_2=-t_2\sum_j(c_{j+2}^\dagger c_j+\mathrm{h.c.}).

    Derive the modified dispersion. Explain how t1t_1 and t2t_2 are related to Fourier components of E(k)E(k). Why does the number and quality of available kk points matter when fitting a model?

Takeaway: For a specified single-orbital basis, real-space hopping amplitudes are encoded in the Fourier structure of the band dispersion.

(b) Reconstruct a two-band model

A second hypothetical calculation produces two bands. For this part, the complete band data are given analytically by

E±(k)=±t12+t22+2t1t2cos⁡(ka),t1,t2>0.E_\pm(k)=\pm\sqrt{t_1^2+t_2^2+2t_1t_2\cos(ka)}, \qquad t_1,t_2>0.

Assume two orbitals, AA and BB, per unit cell, equal on-site energies set to zero, and only nearest-neighbor AA–BB hoppings. A possible Bloch Hamiltonian has the form

H(k)=(0f(k)f(k)∗0).H(k)=\begin{pmatrix}0&f(k)\\f(k)^*&0\end{pmatrix}.
  1. Show that its eigenvalues are E±(k)=±∣f(k)∣E_\pm(k)=\pm|f(k)|. Find one function f(k)f(k) that reproduces the supplied bands.

  2. Write the corresponding Hamiltonian in real space. Identify intracell and intercell hopping terms. Explain why your reconstruction is a possible model, rather than necessarily the unique model.

(c) Can identical bands hide different physics?

Consider two models with positive hopping parameters vv and ww:

HA(k)=(0v+we−ikav+weika0),HB(k)=(0w+ve−ikaw+veika0).H_A(k)=\begin{pmatrix} 0&v+we^{-ika}\\ v+we^{ika}&0 \end{pmatrix}, \qquad H_B(k)=\begin{pmatrix} 0&w+ve^{-ika}\\ w+ve^{ika}&0 \end{pmatrix}.

Assume 0<v<w0<v<w. For all comparisons below, use the same physical orbital positions, the same convention for the unit cell, and open chains consisting of NN complete AA–BB unit cells with no additional end sites.

  1. Same energies? Show that HA(k)H_A(k) and HB(k)H_B(k) have identical bulk energy eigenvalues for every kk.

  2. Different eigenvectors? Find normalized eigenvectors of each model away from band degeneracies. Explain which part of f(k)f(k) is not fixed by the band energies alone.

  3. A geometric distinction. As kk traverses the Brillouin zone, sketch the curves fA(k)=v+we−ikaf_A(k)=v+we^{-ika} and fB(k)=w+ve−ikaf_B(k)=w+ve^{-ika} in the complex plane. Which curve encloses the origin? Determine the winding number of each curve, defining positive winding as counterclockwise traversal. (Hint: Both curves are circles.)

  4. Boundary test. Write the real-space hopping patterns for both models with open boundaries. In the limit v=0v=0 and w>0w>0, identify any uncoupled end sites and their energies. What does this suggest about boundary-localized states when 0<v≪w0<v\ll w? (Optional numerical check: diagonalize chains of increasing NN and examine the near-zero-energy eigenstates.)

  5. Final conceptual question. Two band calculations yield exactly the same En(k)E_n(k). List at least two additional pieces of information that could help distinguish their effective Hamiltonians or physical boundary responses. Explain why those quantities contain information absent from the band energies.

Important convention: The winding number of an SSH Bloch Hamiltonian depends on how the orbital basis and unit cell are chosen. A shift of unit-cell convention alone is not a new physical material. In Question 9, the specified orbital embedding and boundary termination are essential for making a meaningful comparison.

Inspiring question: If a band structure does not uniquely determine the Hamiltonian, what exactly have we learned when we “fit a tight-binding model” to a first-principles calculation?


Final reflection — Two directions, one lesson

In Problem 1, you started from orbitals and symmetry to constrain a Hamiltonian. In Problem 2, you started from energy bands and tried to reconstruct a Hamiltonian.

In a short paragraph (approximately 100–150 words), answer:

Which direction contains more information: constructing a Hamiltonian from orbitals and symmetry, or reconstructing it from energies? What additional assumptions or measurements would make the inverse problem more reliable?

The goal is not merely to calculate energy eigenvalues. It is to understand what a Hamiltonian tells us about a material—and what its spectrum leaves hidden.