Band Structure of Solids
CSIR NET chemistry: bands from MOs, Bloch theorem, free-electron Fermi energy, band gaps, metals vs semiconductors vs insulators, doping, direct/indirect gaps.
- 20 min
On this page · 18 sections
- Prerequisites
- The Hook: Same Group, Different Worlds
- From Molecular Orbitals to Bands
- Bloch’s Theorem and k
- The Free-Electron Model of a Metal
- The Nearly-Free-Electron Model: Where Gaps Come From
- Metals, Semiconductors and Insulators
- Intrinsic Semiconductors
- Extrinsic (Doped) Semiconductors
- Direct and Indirect Gaps
- Where Simple Band Theory Fails
- Memory Tricks
- Don’t Confuse
- Common Mistakes
- Practice
- Quick Revision Table
- Key Takeaways
- What’s Next
Prerequisites
- Hückel theory for linear and cyclic polyenes from Hückel Theory for Conjugated π Systems. A band is what Hückel’s MO ladder becomes when N is about \( 10^{23} \).
- The particle in a box from Particle in a Box, Harmonic Oscillator and Tunnelling, which is the free-electron model of a metal.
- Bragg’s law and the reciprocal lattice from Bragg’s Law and X-ray Diffraction. Band gaps open exactly where electron waves satisfy the Bragg condition.
The Hook: Same Group, Different Worlds
From Molecular Orbitals to Bands
Take a chain of N identical atoms, each contributing one s orbital, and apply Hückel theory with Coulomb integral α and resonance integral β (negative). For an open chain the N MOs have energies
$$E_j = \alpha + 2\beta\cos\left(\frac{j\pi}{N + 1}\right), \qquad j = 1, 2, \ldots, N$$As N grows the levels stay between \( \alpha + 2\beta \) and \( \alpha - 2\beta \) but get closer together. For \( N \approx 10^{23} \) they form a practically continuous band of width
$$\boxed{W = 4|\beta|}$$graph LR
A["1 atom:
one level α"] --> B["2 atoms:
α ± β"]
B --> C["4 atoms:
4 levels,
spread < 4|β|"]
C --> D["N atoms:
a band of width 4|β|,
N levels"]Four conclusions follow, and they carry over to real 3D solids:
- A band holds 2N electrons (N orbitals, two spins each).
- Band width depends on overlap. Strongly overlapping valence s and p orbitals give wide bands (several eV); compact d orbitals give narrower bands; 4f orbitals barely overlap and give very narrow bands, so lanthanide ions keep atomic-like properties in solids.
- Bonding at the bottom, antibonding at the top. The lowest level is the all-in-phase combination; the highest alternates in sign.
- Each atomic orbital type gives its own band (an s band, a p band, a d band). They may be separated by a gap or may overlap.
Density of states
The density of states g(E) counts levels per unit energy. It is not uniform. In the 1D Hückel chain levels crowd at the top and bottom of the band. In a 3D free-electron metal \( g(E) \propto \sqrt E \). Real bands have a complicated g(E), but the idea that the number of states in a band is fixed (2N electrons) is what decides metal versus insulator.
Bloch’s Theorem and k
In a periodic potential with period a, the one-electron wavefunctions take the Bloch form
$$\boxed{\psi_k(x) = e^{ikx}\,u_k(x), \qquad u_k(x + a) = u_k(x)}$$a plane wave modulated by a function with the period of the lattice. The label k is the crystal wave vector. For the cyclic Hückel chain (a ring of N atoms, the model for an infinite crystal with periodic boundary conditions):
$$E(k) = \alpha + 2\beta\cos(ka), \qquad k = \frac{2\pi m}{Na}$$All distinct states lie in the range \( -\pi/a < k \le \pi/a \), the first Brillouin zone; values of k outside it repeat states inside. There are exactly N allowed k values in the zone, one per unit cell, which is the same counting as before.
Effective mass. Near the bottom of the band (k = 0, since β < 0), expanding the cosine gives \( E \approx \alpha + 2\beta + |\beta|a^2k^2 \). Comparing with the free-electron \( E = \hbar^2k^2/2m \) defines an effective mass:
$$m^* = \frac{\hbar^2}{d^2E/dk^2} = \frac{\hbar^2}{2|\beta|a^2}$$Narrow bands (small |β|) mean heavy, sluggish carriers; wide bands mean light, mobile ones. Near the top of a band \( d^2E/dk^2 < 0 \), so electrons there have negative effective mass. It is simpler to describe the few empty states at the top of an almost full band as holes: positive charge, positive effective mass.
The Free-Electron Model of a Metal
The opposite starting point: ignore the ions altogether and treat valence electrons as particles in a 3D box of volume V. The energy is \( E = \hbar^2k^2/2m_e \), and filling the lowest states with n electrons per unit volume (two per state) up to the Fermi energy gives
$$\boxed{E_F = \frac{\hbar^2}{2m_e}\left(3\pi^2 n\right)^{2/3}}$$Worked example: copper. One conduction electron per atom, fcc with \( a = 361.5 \) pm, so \( n = 4/a^3 = 8.47 \times 10^{28} \) m⁻³:
$$E_F = \frac{(1.055 \times 10^{-34})^2}{2 \times 9.109 \times 10^{-31}}\left(3\pi^2 \times 8.47 \times 10^{28}\right)^{2/3} = 1.13 \times 10^{-18}\ \text{J} = 7.0\ \text{eV}$$The Fermi temperature \( T_F = E_F/k_B \approx 8 \times 10^4 \) K. Because \( T_F \gg \) room temperature, only electrons within about \( k_BT \) of \( E_F \) can be excited. That explains a classic puzzle: the electronic heat capacity of a metal is far smaller than the classical \( \tfrac32 R \) per mole of electrons, because only a fraction of order \( T/T_F \) of the electrons can take up thermal energy. The average energy of the free-electron gas at 0 K is \( \tfrac35E_F \).
The Fermi–Dirac distribution
At temperature T the probability that a state of energy E is occupied is
$$f(E) = \frac{1}{e^{(E - E_F)/k_BT} + 1}$$- At \( E = E_F \), \( f = \tfrac12 \) at any temperature: this is the working definition of the Fermi level.
- At 0 K, f is a step: 1 below \( E_F \), 0 above.
- For \( E - E_F \gg k_BT \), f reduces to the Boltzmann factor \( e^{-(E - E_F)/k_BT} \). Example: at 300 K (\( k_BT = 0.0259 \) eV), a state 0.10 eV above \( E_F \) is occupied with probability \( 1/(e^{3.87} + 1) = 0.020 \).
The Nearly-Free-Electron Model: Where Gaps Come From
Now switch on a weak periodic potential from the ions. Most free-electron states hardly change, except those whose wavelength fits the lattice. An electron wave travelling along a row of atoms with spacing a is Bragg reflected when
$$2a = n\lambda \quad\Longleftrightarrow\quad k = \frac{2\pi}{\lambda} = \frac{n\pi}{a}$$This is Bragg’s law at normal incidence (θ = 90°, d = a). At \( k = \pm\pi/a \), the forward and reflected waves combine into two standing waves:
| Standing wave | Electron density | Energy |
|---|---|---|
| \( \psi_+ \propto \cos(\pi x/a) \) | piled up on the positive ions | lowered |
| \( \psi_- \propto \sin(\pi x/a) \) | piled up between the ions | raised |
The two have different energies at the same k, so a band gap opens at the Brillouin zone boundary, of size \( E_g = 2|U_1| \), where \( U_1 \) is the Fourier component of the lattice potential. The tight-binding (Hückel) picture and the nearly-free-electron picture are two ends of the same physics: one starts from atoms and widens levels into bands, the other starts from free electrons and cuts gaps into the parabola.
Metals, Semiconductors and Insulators
| Class | Band picture | \( E_g \) | Conductivity versus T |
|---|---|---|---|
| Metal | a partly filled band, or overlapping bands | none | decreases as T rises (more phonon scattering) |
| Semimetal | a tiny overlap of bands | about zero | small carrier density; weak T dependence |
| Semiconductor | full valence band, empty conduction band, small gap | roughly 0.1 to 3 eV | increases steeply with T (more carriers) |
| Insulator | the same, with a large gap | several eV | negligible |
The boundary between “semiconductor” and “insulator” is a matter of convention, not a sharp physical line.
Counting decides. Since each band holds 2 electrons per primitive cell:
- Alkali metals (one valence electron per atom, one atom per primitive cell): the s band is half full. Metal.
- Alkaline earth metals (two electrons): an isolated s band would be exactly full, which would make Be or Mg an insulator. They are metals because the s and p bands overlap, so electrons spill into both partly filled bands.
- Diamond, Si, Ge (four electrons, two atoms per primitive cell, eight electrons): sp³-type bonding and antibonding bands separate, the bonding (valence) band is exactly full, and a gap remains.
- Graphite: in a single layer (graphene) the π and π* bands just touch at the corners of the zone; in graphite, interlayer interaction makes them overlap slightly. It is a semimetal, and conducts well within the layers but poorly between them.
Band gaps worth knowing
| Material | \( E_g \) at about 300 K (eV) | Type of gap | Absorption edge \( \lambda = 1240/E_g \) nm |
|---|---|---|---|
| Ge | 0.66 | indirect | about 1880 nm |
| Si | 1.12 | indirect | about 1110 nm |
| GaAs | 1.42 | direct | about 870 nm |
| CdS | about 2.4 | direct | about 515 nm |
| ZnO | about 3.3 | direct | about 375 nm |
| Diamond | about 5.5 | indirect | about 225 nm |
A semiconductor absorbs every photon with \( \lambda < 1240/E_g \) and transmits longer wavelengths.
This explains the colours of semiconductors. CdS absorbs violet and blue (below about 515 nm) and looks yellow. ZnO absorbs only in the UV and looks white, which is why it is used in sunscreens. Silicon absorbs all visible light and looks grey-black. Diamond transmits all visible light and is colourless.
Trends in the gap
- Down group 14 the gap falls as overlap between larger orbitals widens the bands: C > Si > Ge > grey Sn (essentially zero) and Pb is a metal.
- Along an isoelectronic series the gap rises with ionic character: Ge (0.66 eV) < GaAs (1.42 eV) < ZnSe (about 2.7 eV). The more electronegative partner pulls the valence band down.
- Small particles of a semiconductor (quantum dots a few nanometres across) have larger gaps than the bulk, because the electron and hole are confined, exactly as in a particle in a box. Their colour shifts towards blue as the particles shrink.
Intrinsic Semiconductors
In a pure semiconductor at T > 0, thermal excitation lifts electrons across the gap, leaving an equal number of holes: \( n = p = n_i \). Statistical mechanics gives
$$\boxed{n_i = \sqrt{N_cN_v}\,\exp\left(-\frac{E_g}{2k_BT}\right), \qquad N_c, N_v \propto T^{3/2}}$$The factor of 2 in the exponent arises because each excitation creates two carriers (the same reason as for Schottky defects on the ionic structures page). The conductivity is
$$\sigma = e\,(n\mu_e + p\mu_h)$$with mobilities μ. Because the exponential dominates the weak \( T^{3/2} \) prefactor and the slow change in mobility,
$$\ln\sigma \approx \text{constant} - \frac{E_g}{2k_BT}$$A plot of \( \ln\sigma \) against \( 1/T \) has slope \( -E_g/2k_B \). The Fermi level of an intrinsic semiconductor lies near the middle of the gap: \( E_F = \tfrac12(E_c + E_v) + \tfrac34k_BT\ln(m_h^*/m_e^*) \).
Worked example. Ignoring the prefactor, the intrinsic carrier density of Ge compared with Si at 300 K is \( \exp[(1.12 - 0.66)/(2 \times 0.02585)] = e^{8.9} \approx 7 \times 10^3 \) times larger. A small difference in gap means thousands of times more carriers.
Extrinsic (Doped) Semiconductors
Replacing a tiny fraction of silicon atoms with a neighbouring-group element adds carriers without heating:
| Dopant in Si or Ge | Group | Extra | Level in the gap | Majority carrier | Type |
|---|---|---|---|---|---|
| P, As, Sb | 15 | one electron | donor level just below the conduction band | electrons | n-type |
| B, Al, Ga, In | 13 | one missing electron | acceptor level just above the valence band | holes | p-type |
The donor electron is bound to the P⁺ ion only weakly: in a hydrogen-like model the binding energy is \( 13.6\ \text{eV} \times (m^*/m_e)/\varepsilon_r^2 \), which the large dielectric constant of silicon reduces to a few tens of meV, comparable to \( k_BT \) at room temperature. So almost every donor is ionised at room temperature.
Example. Si with \( n_i = 1.0 \times 10^{10} \) cm⁻³ doped with \( N_D = 1.0 \times 10^{16} \) cm⁻³ P atoms: \( n = 10^{16} \), \( p = 10^{20}/10^{16} = 10^4 \) cm⁻³. With an electron mobility of 1350 cm² V⁻¹ s⁻¹, \( \sigma = 10^{16} \times 1.602 \times 10^{-19} \times 1350 = 2.2 \) S cm⁻¹, enormously higher than intrinsic silicon.
The Fermi level moves towards the conduction band in n-type material and towards the valence band in p-type. A doped semiconductor shows three temperature regimes: at very low T carriers freeze back onto dopants (freeze-out); over a middle range all dopants are ionised and n is nearly constant (extrinsic or saturation regime); at high T intrinsic carriers swamp the dopants (intrinsic regime). The sign of the Hall coefficient \( R_H = 1/nq \) tells which carrier dominates.
Doping by non-stoichiometry
In ionic oxides, defects do the job of dopants (see Close Packing and Ionic Crystal Structures):
- Metal excess (ZnO heated, losing oxygen; TiO₂₋ₓ): extra electrons, n-type.
- Metal deficiency (Ni₁₋ₓO, Fe₁₋ₓO, Cu₂₋ₓO): higher oxidation states act as holes, p-type.
- Valence induction: NiO doped with Li₂O, \( \text{Li}^+_x\text{Ni}^{2+}_{1-2x}\text{Ni}^{3+}_x\text{O} \), becomes a much better p-type conductor.
Direct and Indirect Gaps
The minimum of the conduction band and the maximum of the valence band need not occur at the same k.
| Direct gap | Indirect gap | |
|---|---|---|
| Band extrema | at the same k | at different k |
| Photon absorption or emission at the edge | allowed directly (a photon carries almost no momentum) | needs a phonon to supply the momentum; weak |
| Light emission | efficient | very inefficient |
| Examples | GaAs, GaN, InP, CdS, ZnO | Si, Ge, diamond, GaP |
| Use | LEDs, laser diodes | electronics, solar cells (thicker absorbers needed) |
This is why LEDs are built from III–V and II–VI compounds and not from silicon.
Where Simple Band Theory Fails
Part C sometimes asks for the limits of the model.
- Mott insulators. NiO and CoO have partly filled d bands and should be metals by band counting, yet they are insulators. Strong repulsion between d electrons on the same ion (electron correlation) splits the band. Simple one-electron band theory leaves this out. By contrast, TiO and ReO₃, where the d orbitals overlap well through the lattice, are metallic.
- Peierls distortion. A 1D chain with a half-filled band is unstable: it dimerises into alternating short and long bonds, which doubles the cell and opens a gap at the Fermi level. Polyacetylene, with alternating single and double bonds, is the chemical example; it becomes highly conducting only when doped.
- Narrow f bands. 4f electrons stay localised, which is why lanthanide magnetism and spectra in solids look atomic.
Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| Fermi level (f = ½) | Fermi energy of a metal at 0 K (top filled level) | |
| Intrinsic: \( n = p \), \( E_F \) mid-gap | extrinsic: one carrier dominates, \( E_F \) near a band edge | |
| n-type: donors, electrons majority | p-type: acceptors, holes majority; both are electrically neutral overall | |
| \( \exp(-E_g/2kT) \) for carriers | \( \exp(-E_g/kT) \), the common mistake | |
| Direct gap (GaAs) | indirect gap (Si, Ge, diamond) | |
| Metal resistivity rises with T | semiconductor resistivity falls with T | |
| Semimetal (graphite, grey Sn) | metal (partly filled band, many carriers) | |
| Band insulator (diamond) | Mott insulator (NiO), partly filled d band but insulating |
Common Mistakes
Practice
All questions below are original practice questions written for this page. They are not from any CSIR NET paper.
Level 1: Foundation
- What is the occupation probability of a state at the Fermi level at any temperature above 0 K?
- Silicon doped with arsenic is n-type or p-type?
- Name one direct-gap and one indirect-gap semiconductor.
Answers
- One half.
- n-type (As is group 15, a donor).
- Direct: GaAs (or GaN, CdS, ZnO). Indirect: Si (or Ge, diamond).
Level 2: Part B Style
Practice Q1. A semiconductor has a band gap of 2.42 eV. Its absorption edge and appearance are best described as:
(a) 1107 nm; black (b) 512 nm; yellow (c) 368 nm; white (d) 873 nm; dark grey
Answer: (b)
\( \lambda = 1240/2.42 = 512 \) nm. Light shorter than 512 nm (violet, blue, some green) is absorbed; the transmitted and reflected light looks yellow. These are the properties of CdS.
Practice Q2. In the Hückel model of an infinite chain of atoms with one s orbital each, the width of the band formed is:
(a) |β| (b) 2|β| (c) 4|β| (d) N|β|
Answer: (c) 4|β|
\( E = \alpha + 2\beta\cos(ka) \) runs from \( \alpha + 2\beta \) to \( \alpha - 2\beta \). The width is independent of N once N is large; N sets only the number of levels.
Practice Q3. Silicon with \( n_i = 1.0 \times 10^{10} \) cm⁻³ is doped with \( 1.0 \times 10^{16} \) cm⁻³ boron atoms. At room temperature the electron concentration is about:
(a) \( 10^{16} \) cm⁻³ (b) \( 10^{10} \) cm⁻³ (c) \( 10^{4} \) cm⁻³ (d) zero
Answer: (c) 10⁴ cm⁻³
Boron is an acceptor, so \( p \approx 10^{16} \) cm⁻³ and \( n = n_i^2/p = 10^{20}/10^{16} = 10^4 \) cm⁻³. The minority carrier density is small but not zero.
Practice Q4. Which statement about the temperature dependence of electrical conductivity is correct?
(a) It increases with temperature for both copper and silicon. (b) It decreases with temperature for copper and increases for intrinsic silicon. (c) It increases for copper and decreases for silicon. (d) It is independent of temperature for both.
Answer: (b)
In copper the carrier density is fixed and lattice vibrations scatter electrons more at higher T. In intrinsic silicon the number of carriers rises exponentially with T.
Level 3: Part C Style
Practice Q5. The conductivity of an intrinsic semiconductor rises by a factor of 20.9 when it is heated from 300 K to 350 K. Assume the mobilities and the \( T^{3/2} \) prefactor are constant (\( k_B = 8.617 \times 10^{-5} \) eV K⁻¹). Which statements are correct?
- The band gap is about 1.10 eV.
- The band gap is about 0.55 eV.
- Including the \( T^{3/2} \) prefactor would make the estimated gap slightly smaller.
- The material absorbs light at all wavelengths shorter than about 1130 nm.
(a) 1 and 3 only (b) 1, 3 and 4 only (c) 2 and 4 only (d) 2, 3 and 4 only
Answer: (b) 1, 3 and 4 only
\( \ln 20.9 = \dfrac{E_g}{2k_B}\left(\dfrac{1}{300} - \dfrac{1}{350}\right) \Rightarrow E_g = \dfrac{2 \times 8.617 \times 10^{-5} \times 3.040}{4.762 \times 10^{-4}} = 1.10 \) eV. Statement 1 correct; 2 is the result of forgetting the factor 2.
Part of the factor 20.9 is then explained by \( (350/300)^{3/2} = 1.26 \), leaving \( 20.9/1.26 = 16.6 \) for the exponential, which gives \( E_g \approx 1.02 \) eV. Correct.
\( 1240/1.10 = 1127 \) nm. Correct.
Practice Q6. Treat silver (fcc, \( a = 408.6 \) pm, one conduction electron per atom) as a free-electron metal. Which statements are correct?
- The conduction electron density is about \( 5.9 \times 10^{28} \) m⁻³.
- The Fermi energy is about 5.5 eV.
- At 300 K, most conduction electrons can absorb thermal energy of order \( k_BT \).
- A state 0.10 eV above \( E_F \) at 300 K is occupied with probability about 2%.
(a) 1 and 2 only (b) 1, 2 and 4 only (c) 2, 3 and 4 only (d) 1, 2, 3 and 4
Answer: (b) 1, 2 and 4 only
- \( n = 4/a^3 = 4/(4.086 \times 10^{-10})^3 = 5.86 \times 10^{28} \) m⁻³. Correct.
- \( E_F = \dfrac{\hbar^2}{2m_e}(3\pi^2n)^{2/3} = 8.8 \times 10^{-19} \) J = 5.5 eV. Correct.
- Wrong. Only electrons within about \( k_BT \) (0.026 eV) of \( E_F \) (5.5 eV) can be excited: a fraction of order \( T/T_F \approx 0.005 \). This is why the electronic heat capacity is small.
- \( f = 1/(e^{0.10/0.02585} + 1) = 1/(47.9 + 1) = 0.020 \). Correct.
Practice Q7. Which statements are consistent with band theory and its known limitations?
- Magnesium is metallic because its 3s and 3p bands overlap.
- NiO is an insulator although simple band counting predicts a partly filled d band; electron correlation is responsible.
- Silicon is a poor light emitter because its band gap is indirect.
- A one-dimensional chain with a half-filled band is stable against dimerisation.
(a) 1 and 2 only (b) 1, 2 and 3 only (c) 2, 3 and 4 only (d) 1, 2, 3 and 4
Answer: (b) 1, 2 and 3 only
1, 2 and 3 are correct as explained in the text. 4 is wrong: the Peierls theorem says the half-filled 1D chain lowers its energy by dimerising, which opens a gap at the Fermi level (polyacetylene).
Quick Revision Table
| Idea | Formula or fact |
|---|---|
| Hückel chain | \( E_j = \alpha + 2\beta\cos[j\pi/(N + 1)] \); width 4\( \vert\beta\vert \) |
| Bloch | \( \psi_k = e^{ikx}u_k(x) \); first zone \( -\pi/a < k \le \pi/a \) |
| Band capacity | 2N electrons (2 per primitive cell) |
| Effective mass | \( m^* = \hbar^2/(d^2E/dk^2) \); narrow band, heavy carrier |
| Free electrons | \( E_F = (\hbar^2/2m)(3\pi^2n)^{2/3} \); Cu about 7.0 eV; \( g(E) \propto \sqrt E \) |
| Fermi–Dirac | \( f = 1/[e^{(E - E_F)/kT} + 1] \); \( f(E_F) = \tfrac12 \) |
| Gap origin | Bragg reflection at \( k = n\pi/a \); \( E_g = 2\vert U\vert \) |
| Gaps (eV) | Ge 0.66, Si 1.12, GaAs 1.42, CdS 2.4, ZnO 3.3, diamond 5.5 |
| Edge | \( \lambda = 1240/E_g \) nm |
| Intrinsic | \( n_i \propto T^{3/2}e^{-E_g/2kT} \); slope of ln σ vs 1/T is \( -E_g/2k \) |
| Doping | group 15 donor n-type; group 13 acceptor p-type; \( np = n_i^2 \) |
| Oxides | ZnO (excess Zn) n-type; NiO, FeO (deficient) p-type |
| Direct / indirect | GaAs, GaN, CdS / Si, Ge, diamond |
| T dependence | metal σ falls; semiconductor σ rises |
| Limits | Mott (NiO), Peierls (1D chains), narrow 4f bands |
Key Takeaways
What’s Next
- Solid State: Quick Revision: every formula and trap from this chapter on one page.
- For the Hückel algebra behind bands: Hückel Theory for Conjugated π Systems.
- Electronic transitions and absorption edges in molecules: Molecular Spectroscopy.
- Back to the chapter overview.