Advanced Band Structure Methods & Many-Body Physics in Solids
This culminating unit covers advanced computational and many-body theoretical techniques in modern solid state physics. We analyze electron-electron interactions, Coulomb exchange energy, and the failure of single-particle Hartree-Fock theory at the Fermi surface. We derive the Lindhard dielectric response function, Thomas-Fermi screening, Kohn anomalies, and Friedel oscillations. We explore Fermi surface topology, de Haas-van Alphen quantum oscillations, and Wannier functions. We study modern electronic structure calculation methods including tight-binding, pseudopotentials, and Density Functional Theory (DFT). Finally, we explore strongly correlated electrons within the Hubbard model, the Mott metal-insulator transition, the GW approximation, and semiconductor superlattices.
§8.1 Electron-Electron Interactions: The Hartree & Hartree-Fock Approximations in the Electron Gas
1. The Many-Body Hamiltonian & Jellium Model
In real solids, electrons interact not only with the periodic ionic lattice potential $V_{\text{ion}}(\vec{r})$, but strongly with each other via the long-range Coulomb repulsion. In the uniform electron gas (jellium) model:
$$\hat{H} = \sum_{i=1}^N \frac{p_i^2}{2m} + \frac{1}{2}\sum_{i \ne j} \frac{e^2}{4\pi\epsilon_0 |\vec{r}_i - \vec{r}_j|} + \hat{H}_{\text{bg}} + \hat{H}_{\text{e-bg}}$$The positive ion background $\hat{H}_{\text{bg}}$ and electron-background interaction $\hat{H}_{\text{e-bg}}$ exactly cancel the direct (Hartree) classical electrostatic divergence at $\vec{q} = 0$.
2. The Hartree-Fock Exchange Energy
In the Hartree-Fock approximation, the $N$-electron wavefunction is represented by a single Slater determinant of plane waves. The total energy per electron is:
$$E_{\text{HF}} / N = \langle \hat{T} \rangle + E_{\text{exchange}} = \frac{3}{5} E_F - \frac{3}{4} \frac{e^2 k_F}{4\pi^2 \epsilon_0}$$Introducing the dimensionless Wigner-Seitz density parameter $r_s \equiv r_0 / a_0$ (where $\frac{4}{3}\pi r_0^3 = 1/n$):
$$\frac{E_{\text{HF}}}{N} = \left[ \frac{2.21}{r_s^2} - \frac{0.916}{r_s} \right] \text{ Rydbergs}$$- At high densities ($r_s \ll 1$): Kinetic energy ($\propto 1/r_s^2$) dominates $\implies$ Free electron gas behavior.
- At low densities ($r_s \gg 1$): Exchange and correlation energies ($\propto -1/r_s$) dominate $\implies$ Wigner electron crystallization!
Failure of Hartree-Fock at the Fermi Surface: The single-particle Hartree-Fock energy dispersion is: $$\epsilon_{\text{HF}}(k) = \frac{\hbar^2 k^2}{2m} - \frac{e^2 k_F}{2\pi^2 \epsilon_0} \left[ 1 + \frac{k_F^2 - k^2}{2 k k_F} \ln\left| \frac{k + k_F}{k - k_F} \right| \right]$$ Differentiating to obtain the group velocity $v_g = \frac{1}{\hbar}\frac{d\epsilon}{dk}$, the logarithmic term causes the velocity to diverge logarithmically to infinity at the Fermi surface ($k = k_F$), forcing the density of states $g(E_F) \propto 1/v_g$ to drop unphysically to zero! This catastrophe occurs because Hartree-Fock completely neglects dynamic dielectric screening of the Coulomb interaction by other electrons.
§8.2 Linear Dielectric Response Theory, Lindhard Function & Thomas-Fermi Screening
1. Linear Dielectric Response Formalism
When an external test charge $\rho_{\text{ext}}(\vec{r}, t) = \rho_0 e^{i(\vec{q}\cdot\vec{r} - \omega t)}$ is introduced into the electron gas, it induces a screening charge density $\rho_{\text{ind}}(\vec{q}, \omega)$. In linear response theory:
$$\rho_{\text{ind}}(\vec{q}, \omega) = \chi(\vec{q}, \omega) \phi_{\text{total}}(\vec{q}, \omega)$$where $\chi(\vec{q}, \omega)$ is the dynamic polarizability. The frequency- and wavevector-dependent dielectric function $\epsilon(\vec{q}, \omega)$ is defined by:
$$\phi_{\text{total}}(\vec{q}, \omega) = \frac{\phi_{\text{ext}}(\vec{q}, \omega)}{\epsilon(\vec{q}, \omega)} \implies \epsilon(\vec{q}, \omega) = 1 - \frac{e^2}{\epsilon_0 q^2} \chi(\vec{q}, \omega)$$2. The Lindhard Dielectric Function
Jens Lindhard (1954) derived the exact quantum polarizability of a degenerate Fermi gas in the Random Phase Approximation (RPA):
$$\epsilon(\vec{q}, \omega) = 1 + \frac{e^2}{\epsilon_0 q^2} \sum_{\vec{k}} \frac{f(\vec{k}) - f(\vec{k}+\vec{q})}{\epsilon(\vec{k}+\vec{q}) - \epsilon(\vec{k}) - \hbar\omega - i\eta}$$In the static limit ($\omega = 0$), the static Lindhard dielectric function is:
$$\epsilon(q, 0) = 1 + \frac{q_{\text{TF}}^2}{q^2} F\left( \frac{q}{2k_F} \right)$$where $q_{\text{TF}} = \sqrt{\frac{3 e^2 n}{2\epsilon_0 E_F}}$ is the Thomas-Fermi screening wavevector, and $F(x)$ is the Lindhard function ($x \equiv q / 2k_F$):
$$F(x) = \frac{1}{2} + \frac{1 - x^2}{4x} \ln\left| \frac{1 + x}{1 - x} \right|$$- Long-Wavelength Limit ($q \ll 2k_F$, $x \to 0$): $F(x) \to 1$, giving the classical Thomas-Fermi screening: $$\epsilon(q) \approx 1 + \frac{q_{\text{TF}}^2}{q^2} \implies V(r) = \frac{e}{4\pi\epsilon_0 r} \exp(-q_{\text{TF}} r)$$ The Coulomb potential is screened exponentially over the Thomas-Fermi length $\lambda_{\text{TF}} = 1/q_{\text{TF}} \sim 0.5\text{ \AA}$.
- Zone Edge Singularity ($q = 2k_F$, $x = 1$): The derivative $dF/dx$ diverges logarithmically to $-\infty$ due to the sharp step in the Fermi-Dirac distribution at $T = 0\text{ K}$.
§8.3 Kohn Anomalies, Friedel Oscillations & The Friedel Sum Rule
1. Real-Space Friedel Oscillations
Because the static Lindhard dielectric function $\epsilon(q, 0)$ possesses a non-analytic logarithmic derivative at $q = 2k_F$, taking the inverse Fourier transform of the screened potential $\phi(r) = \int \frac{\phi_{\text{ext}}(q)}{\epsilon(q)} e^{i\vec{q}\cdot\vec{r}} \frac{d^3q}{(2\pi)^3}$ produces long-range spatial ripples rather than pure exponential decay:
$$\delta\rho(r) \propto \frac{\cos(2k_F r)}{r^3} \quad (r \gg 1/k_F)$$ $$\phi(r) \propto \frac{\cos(2k_F r)}{r^3}$$These quantum interference ripples are Friedel oscillations. They arise from the sharp edge of the Fermi sphere: electrons cannot screen perturbations on spatial scales shorter than the Fermi wavelength $\lambda_F = 2\pi/k_F$. Friedel oscillations govern the oscillatory sign of the RKKY magnetic exchange interaction and mediate long-range vacancy-impurity interactions in alloys.
2. Kohn Anomalies & The Friedel Sum Rule
- Kohn Anomaly in Phonon Dispersion: In 1959, Walter Kohn showed that because ions are screened by conduction electrons, the singularity in $\epsilon(q, 0)$ at $q = 2k_F$ is directly imprinted onto the phonon dispersion curve $\omega(\vec{q})$. At phonon wavevectors spanning the Fermi surface ($|\vec{q}| = 2k_F$), the phonon frequency exhibits a sharp dip or kink: $$\left. \frac{d\omega}{dq} \right|_{q = 2k_F} \to -\infty$$ In low-dimensional conductors (1D and 2D), this anomaly triggers the Peierls structural transition and charge density waves (CDWs).
- The Friedel Sum Rule: When an impurity ion with excess nuclear charge $Z e$ is embedded in a metal, it scatters conduction electrons. The phase shifts $\delta_l(E_F)$ of the partial waves at the Fermi energy must completely screen the impurity charge: $$Z = \frac{2}{\pi} \sum_{l=0}^\infty (2l + 1) \delta_l(E_F)$$ This exact relation connects microscopic electron scattering phase shifts to macroscopic residual electrical resistivity $\Delta\rho$.
§8.4 Geometry of the Fermi Surface, Harrison Construction & Quantum Oscillations
1. The Harrison Construction & Fermi Surface Topology
The Fermi surface is the constant-energy locus $E(\vec{k}) = E_F$ separating occupied from unoccupied states in reciprocal space at $T = 0\text{ K}$.
In the nearly free electron approximation, the Harrison construction determines the Fermi surface geometry:
- Draw the free electron Fermi sphere of radius $k_F = (3\pi^2 n)^{1/3}$ centered at every reciprocal lattice point $\vec{G}$.
- Points enclosed by at least one sphere belong to the first Brillouin zone; points enclosed by at least two spheres belong to the second zone, and so on.
- Applying periodic potential gaps rounds off the sharp intersections, decomposing the Fermi surface into:
- Closed Electron Pockets: Enclosing lower unoccupied states.
- Closed Hole Pockets: Enclosing unoccupied states.
- Open Orbits: Non-periodic trajectories that traverse the entire Brillouin zone without closing, producing unsaturated transverse magnetoresistance ($\rho_{xx} \propto B^2$).
2. The de Haas-van Alphen (dHvA) Effect & Onsager Relation
When a strong magnetic field $B\hat{z}$ is applied to a clean metal at cryogenic temperatures ($\omega_c \tau \gg 1, k_B T \ll \hbar\omega_c$), the continuous electronic density of states condenses into discrete Landau tubes in reciprocal space with cross-sectional area:
$$S_n = (n + \gamma) \frac{2\pi e B}{\hbar}$$As $B$ increases, the Landau tubes expand radially. Each time a tube crosses the extremal cross-section of the Fermi surface $S_e$, the free energy and magnetic moment undergo a sharp oscillation.
According to the Onsager Relation (1952), the magnetic susceptibility oscillates periodically in $1/B$:
$$\Delta\left( \frac{1}{B} \right) = \frac{2\pi e}{\hbar S_e}$$By measuring the oscillation periods $\Delta(1/B)$ as a function of the magnetic field orientation, the complete 3D topological shape and extremal cross-sectional areas $S_e$ of the Fermi surface can be reconstructed with sub-percent precision.
§8.5 Modern Computational Band Structure: Tight-Binding, Pseudopotentials & Density Functional Theory
1. The Tight-Binding Method
In the tight-binding (LCAO) approximation, the crystal wavefunction is constructed from linear combinations of isolated atomic orbitals $\phi_m(\vec{r})$:
$$\psi_{\vec{k}}(\vec{r}) = \frac{1}{\sqrt{N}} \sum_j e^{i \vec{k} \cdot \vec{R}_j} \phi(\vec{r} - \vec{R}_j)$$Retaining only on-site energy $\epsilon_0$ and nearest-neighbor hopping matrix element $t = -\int \phi^*(\vec{r}) \Delta V(\vec{r}) \phi(\vec{r} - \vec{\delta}) d^3r$, the band dispersion on a 2D square lattice of constant $a$ is:
$$E(\vec{k}) = \epsilon_0 - 2t [\cos(k_x a) + \cos(k_y a)]$$At the band saddle points ($\vec{k} = (\pi/a, 0)$ and $(0, \pi/a)$), the group velocity vanishes, producing a logarithmic divergence in the 2D density of states known as a van Hove singularity:
$$g(E) \propto \ln\left| \frac{16t}{E - \epsilon_0} \right|$$2. Pseudopotentials & Density Functional Theory (DFT)
- Orthogonalized Plane Waves (OPW) & Pseudopotentials: Conduction wavefunctions must oscillate rapidly near atomic nuclei to remain orthogonal to tightly bound core states, requiring thousands of plane waves. Phillips and Kleinman proved that this core orthogonality acts as an effective repulsive potential that nearly cancels the deep Coulomb attraction: $$V_{\text{pseudo}}(\vec{r}) = V_{\text{true}}(\vec{r}) + \sum_c (E - E_c) |\phi_c\rangle\langle\phi_c|$$ The resulting smooth, shallow pseudopotential reproduces valence eigenvalues accurately with a small plane-wave basis.
- Density Functional Theory (Hohenberg-Kohn & Kohn-Sham, 1964-1965): DFT establishes that the ground-state properties of an interacting many-electron system are uniquely determined by the 3D ground-state electron density $n(\vec{r})$, reducing a $3N$-dimensional problem to 3 dimensions! The Kohn-Sham self-consistent single-particle equations are: $$\left[ -\frac{\hbar^2}{2m} \nabla^2 + V_{\text{ext}}(\vec{r}) + V_{\text{Hartree}}[n](\vec{r}) + V_{\text{xc}}[n](\vec{r}) \right] \psi_i(\vec{r}) = \epsilon_i \psi_i(\vec{r})$$ $$n(\vec{r}) = \sum_{i=1}^N |\psi_i(\vec{r})|^2$$ The exchange-correlation potential $V_{\text{xc}}$ is approximated via the Local Density Approximation (LDA) or Generalized Gradient Approximation (GGA). DFT forms the backbone of modern computational materials physics.
§8.6 Strongly Correlated Electron Systems: The Hubbard Model & The Mott Metal-Insulator Transition
1. Failure of Independent-Electron Band Theory: Mott Insulators
According to conventional Bloch band theory, any crystal with an odd number of valence electrons per unit cell must have a half-filled energy band and should therefore be a metal.
However, materials such as $\text{NiO}$, $\text{CoO}$, and undoped cuprates ($\text{La}_2\text{CuO}_4$) have partially filled $3d$ shells, yet are transparent electrical insulators with optical bandgaps exceeding $4\text{ eV}$!
In 1937, Nevill Mott explained that single-particle band theory completely breaks down when the on-site Coulomb repulsion $U$ between two electrons on the same atom exceeds the kinetic hopping bandwidth $W = 2 z t$.
2. The Single-Band Hubbard Model
The minimal Hamiltonian capturing this competition between kinetic itinerancy and Coulomb localization is the Hubbard Model:
$$\hat{H} = -t \sum_{\langle ij \rangle, \sigma} \left( c_{i\sigma}^\dagger c_{j\sigma} + \text{h.c.} \right) + U \sum_i n_{i\uparrow} n_{i\downarrow}$$- Hopping parameter $t$: Promotes electron delocalization into a metallic band of bandwidth $W = 2 z t$.
- On-site Coulomb repulsion $U$ ($U = \int |\phi(r_1)|^2 \frac{e^2}{r_{12}} |\phi(r_2)|^2$): Energy penalty incurred whenever two electrons with opposite spins occupy the same lattice site.
At half-filling (one electron per atom, $\langle n_i \rangle = 1$):
- Weak Coupling ($U / W \ll 1$): Itinerant metallic Fermi liquid.
- Strong Coupling ($U / W \gg 1$): Double occupancy is strictly suppressed. To conduct electricity, an electron must hop to an already occupied site, costing energy $U$. The single band splits into two sub-bands:
- Lower Hubbard Band (LHB): Completely filled ($E \approx 0$).
- Upper Hubbard Band (UHB): Completely empty ($E \approx U$).
- At second-order perturbation theory in $t/U$, virtual hopping between neighboring singly-occupied sites yields an effective antiferromagnetic Heisenberg exchange interaction: $$J = \frac{4 t^2}{U}$$ explaining why virtually all parent Mott insulators are robust antiferromagnets.
§8.7 Green's Function Many-Body Formalism, GW Approximation & Semiconductor Superlattices
1. Green's Functions & The GW Approximation
Standard DFT notoriously underestimates electronic bandgaps by $30-50\%$ (e.g. predicting $0.5\text{ eV}$ for Si instead of $1.12\text{ eV}$, and predicting zero gap for Mott insulators) because Kohn-Sham eigenvalues are Lagrange multipliers of a non-interacting auxiliary system, not true quasiparticle excitation energies.
To accurately compute true quasiparticle spectra, many-body perturbation theory employs the single-particle Green's function $G(\vec{r}, \vec{r}', \omega)$. The quasiparticle equation is:
$$\left[ -\frac{\hbar^2}{2m} \nabla^2 + V_{\text{ion}}(\vec{r}) + V_{\text{Hartree}}(\vec{r}) \right] \psi_n(\vec{r}) + \int \Sigma(\vec{r}, \vec{r}', \epsilon_n) \psi_n(\vec{r}') d^3r' = \epsilon_n \psi_n(\vec{r})$$where $\Sigma$ is the non-local, energy-dependent self-energy operator. In Lars Hedin's GW approximation, the self-energy is expanded to first order in the dynamically screened Coulomb interaction $W = \epsilon^{-1} v$:
$$\Sigma = i G W$$The $GW$ method accurately reproduces experimental bandgaps across semiconductors and insulators to within $0.1\text{ eV}$.
2. Semiconductor Superlattices & Minibands
In 1970, Leo Esaki and Raphael Tsu proposed fabricating artificial 1D periodic structures by alternating ultra-thin epitaxial layers of two different semiconductors (e.g. $\text{GaAs}$ and $\text{Al}_x\text{Ga}_{1-x}\text{As}$) with period $d \sim 5-20\text{ nm} \gg a_{\text{lattice}}$.
This introduces an artificial periodic potential that folds the host Brillouin zone into a mini-Brillouin zone of width $\pi/d$.
- The conduction band splits into narrow minibands of width $\Delta \sim 10-50\text{ meV}$ separated by minigaps.
- Under an applied electric field $\mathcal{E}$, electrons accelerate to the mini-zone boundary and undergo Bragg reflection, executing periodic real-space Bloch oscillations at frequency $\omega_B = e\mathcal{E}d / \hbar$ without scattering.
- This produces negative differential resistance ($dI/dV < 0$), providing the operational foundation for resonant tunneling diodes (RTDs) and quantum cascade lasers (QCLs).
Honors Examination Worked Problems & Solutions
Rigorous step-by-step mathematical proofs and solutions to university degree examination problems.
An impurity of charge $+Q$ is embedded in a 3D degenerate electron gas with Fermi wavevector $k_F$ and Fermi energy $E_F$.\n\n(a) Using the static Thomas-Fermi approximation $\epsilon(q) = 1 + q_{\text{TF}}^2/q^2$, calculate the real-space electrostatic potential $\phi(r)$ by evaluating the 3D Fourier transform integral.\n(b) In the quantum Lindhard dielectric function, the static polarizability has a logarithmic singularity $\frac{d\chi}{dq} \to -\infty$ at $q = 2k_F$. Prove using contour integration that the asymptotic screening charge density oscillates as $\delta\rho(r) \propto \frac{\cos(2k_F r)}{r^3}$.\n(c) For copper ($k_F = 1.36 \times 10^{10}\text{ m}^{-1}$), calculate the spatial period $\lambda_{\text{Friedel}} = \pi / k_F$ of the Friedel oscillations, and explain why this sets the wavelength of the RKKY magnetic interaction.
(a) Thomas-Fermi Screened Potential: The bare external potential of the point charge $+Q$ in reciprocal space is $\phi_{\text{ext}}(q) = \frac{Q}{\epsilon_0 q^2}$. With the Thomas-Fermi dielectric function $\epsilon(q) = 1 + \frac{q_{\text{TF}}^2}{q^2} = \frac{q^2 + q_{\text{TF}}^2}{q^2}$:
The real-space potential is obtained by 3D inverse Fourier transform:
Evaluating via residue calculus (closing the contour in the upper half-plane with a simple pole at $q = +i q_{\text{TF}}$):
Substituting back:
This is the Yukawa / Thomas-Fermi screened potential.
(b) Derivation of Friedel Oscillations: The exact screening charge density in Fourier space is $\delta\rho(q) = -e \chi(q) \phi(q)$. In the Lindhard function, the derivative $d\epsilon/dq$ diverges logarithmically at $q = 2k_F$ due to the term $(1 - x^2) \ln|\frac{1+x}{1-x}|$ with $x = q/2k_F$. Expanding near $q = 2k_F$ (where $q = 2k_F + \delta$):
Integrating by parts twice transfers derivatives onto $\delta\rho(q)$:
Because the second derivative has a non-analytic branch point pole $\propto \frac{1}{q - 2k_F}$ at $q = 2k_F$:
Therefore, the asymptotic real-space behavior at large distances ($r \gg 1/k_F$) is:
This mathematically establishes the algebraic decay and Friedel oscillations.
(c) Numerical Period for Copper: Given $k_F = 1.36 \times 10^{10}\text{ m}^{-1}$:
The spatial period of the charge oscillations in copper is $2.31\text{ \AA}$ (nearly identical to the nearest-neighbor interatomic distance $a/\sqrt{2} \approx 2.55\text{ \AA}$). Physical consequence: A magnetic impurity spin $\vec{S}_1$ induces an oscillatory spin polarization in the conduction electrons with this exact spatial period $\cos(2k_F r)/r^3$. A second impurity spin $\vec{S}_2$ at distance $r$ aligns parallel or antiparallel depending on whether it sits on a crest or a trough of the Friedel wave, directly dictating the RKKY interaction sign.
Electrons on a 2D square lattice of constant $a$ have the nearest-neighbor tight-binding dispersion:\n
\n\n(a) Determine the bandwidth $W = E_{\max} - E_{\min}$ and the location of the band extrema in the first Brillouin zone.\n(b) Prove that a saddle point exists at $\vec{k} = (\pi/a, 0)$ and $(0, \pi/a)$ at energy $E = 0$, and prove that the 2D density of states $g(E)$ diverges logarithmically as $E \to 0$ (Van Hove singularity).\n(c) Sketch the Fermi surface at half-filling ($E_F = 0$). Prove that the Fermi surface is a perfect square tilted at 45°, and demonstrate that it exhibits perfect nesting with nesting wavevector $\vec{Q} = (\pi/a, \pi/a)$.
(a) Band Extrema and Bandwidth:
- Minimum energy occurs at $k_x = 0, k_y = 0$ ($\Gamma$-point):
- Maximum energy occurs at $k_x = \pm \pi/a, k_y = \pm \pi/a$ ($M$-point):
The total bandwidth is:
(b) Saddle Points & Van Hove Singularity: Compute the group velocity components:
At $\vec{k} = (\pi/a, 0)$ ($X$-point):
The energy is:
Evaluating the second derivatives:
Because the principal curvatures have opposite signs (negative along $k_x$, positive along $k_y$), the point $(\pi/a, 0)$ is a saddle point. Near this saddle point, let $k_x = \pi/a + q_x$ and $k_y = q_y$:
The density of states in 2D is:
Since $|\nabla E| = 2t a^2 \sqrt{q_x^2 + q_y^2}$, integrating over the hyperbolic contours $q_y^2 - q_x^2 = E / t a^2$ yields:
As $E \to 0$, $g(E) \to \infty$ diverges logarithmically. This is the 2D van Hove singularity.
(c) Fermi Surface at Half-Filling & Nesting: At half-filling, there is 1 electron per site, filling half the total zone capacity (since each site can hold 2 electrons). By particle-hole symmetry, $E_F = 0$. The Fermi surface contour equation is:
Using the trigonometric identity $\cos A + \cos B = 2\cos\left(\frac{A+B}{2}\right)\cos\left(\frac{A-B}{2}\right)$:
This yields four straight intersecting line segments connecting the saddle points:
Connecting $(\pi/a, 0) \to (0, \pi/a) \to (-\pi/a, 0) \to (0, -\pi/a) \to (\pi/a, 0)$. This forms a perfect square tilted at 45°! Proof of Perfect Nesting: Consider the nesting wavevector $\vec{Q} = (\pi/a, \pi/a)$. Displace any wavevector $\vec{k}$ on the Fermi surface by $\vec{Q}$:
For any state on the Fermi surface ($E(\vec{k}) = 0$):
The entire flat edge of the Fermi surface translates exactly onto the opposing parallel edge under translation by $\vec{Q}$! This perfect nesting produces a divergence in the static susceptibility $\chi(\vec{Q}) \propto \ln^2(T)$, driving an immediate instability toward an antiferromagnetic spin density wave (SDW) or charge density wave (CDW) at infinitesimal Coulomb repulsion $U > 0$.
Two electrons occupy two adjacent lattice sites $1$ and $2$ described by the two-site Hubbard Hamiltonian:\n
\n\n(a) Enumerate the six basis states of the two-electron Hilbert space, separating them into singly-occupied and doubly-occupied configurations.\n(b) Using second-order degenerate perturbation theory in the strong-coupling limit $U \gg t$, calculate the energy of the spin-triplet ($S = 1$) state and the spin-singlet ($S = 0$) state.\n(c) Prove that the effective spin Hamiltonian is equivalent to an antiferromagnetic Heisenberg interaction $\hat{H}_{\text{spin}} = J \vec{S}_1 \cdot \vec{S}_2 + \text{const}$, and derive the exact formula for the exchange coupling constant $J = 4t^2/U$.
(a) Hilbert Space Basis States: There are $\binom{4}{2} = 6$ total states:
- Singly-occupied states (Energy = $0$ in unperturbed limit $t = 0$):
- Triplet states ($S = 1, S_z = +1, 0, -1$):
- Singlet state ($S = 0, S_z = 0$):
- Doubly-occupied states (Energy = $U$ in unperturbed limit):
(b) Perturbation Energy of Triplet and Singlet:
- For the Triplet States ($|T_+\rangle, |T_0\rangle, |T_-\rangle$):
Both electrons have parallel spins. Hopping one electron to the other site would require placing two electrons with the same spin on the same site ($c_{1\sigma}^\dagger c_{1\sigma}^\dagger = 0$). By the Pauli exclusion principle, this is strictly forbidden! Therefore, the hopping operator $\hat{T}$ gives zero when acting on any triplet state:
The triplet states cannot hop virtually, so their energy remains unperturbed at $E_T = 0$.
- For the Singlet State ($|S\rangle$):
Because the two electrons have opposite spins, hopping is allowed by the Pauli principle:
where $|D_+\rangle = \frac{1}{\sqrt{2}} (|D_1\rangle + |D_2\rangle)$ is the symmetric doubly occupied state. The intermediate state $|D_+\rangle$ has unperturbed energy $U$. By second-order perturbation theory:
Thus, virtual hopping to the doubly occupied states lowers the singlet energy by $-\frac{2t^2}{U}$.
(c) Effective Heisenberg Exchange Constant $J$: The energy difference between singlet and triplet is:
The singlet ($S = 0$, antiparallel spins) is lower in energy than the triplet ($S = 1$) by $\frac{2t^2}{U}$. Recall the spin operator identity:
We map this onto an effective spin Hamiltonian:
Evaluating for the two states:
The difference is:
Equating to the microscopic perturbation result:
Because $J = \frac{4t^2}{U} > 0$, the ground state favors antiparallel spin alignment. This proves that in the strong-coupling limit of the Hubbard model ($U \gg t$), the kinetic suppression of charge fluctuations generates an antiferromagnetic Heisenberg exchange interaction! Virtual electron hopping lowers the kinetic energy of the antiparallel state through quantum mechanical hybridization with high-energy polar states.