Band Theory: Insulators, Semiconductors & Metals

Band theory explains why solids are insulators, semiconductors, or metals based on the energy bands formed by electrons in a periodic crystal lattice. Key result: the existence and size of an energy gap between the highest filled band (valence band) and the lowest empty band (conduction band), together with whether the highest occupied band is full or partly filled, determines electrical conductivity. Applies to: crystalline solids, semiconductor devices, electronic structure. Who/when: Felix Bloch (1928), Alan Wilson (1931).

Band theory describes the allowed energy levels of electrons in a crystalline solid. Because atoms are arranged periodically, the discrete energy levels of isolated atoms broaden into continuous bands of allowed energies separated by forbidden gaps. Whether a material conducts electricity depends on two things: the width of the energy gap above the highest filled states, and whether the topmost occupied band is completely full or only partially filled. The surprising fact is that the distinction between a copper wire (metal) and diamond (insulator) — a difference in conductivity of more than twenty orders of magnitude — emerges from the same Schrödinger equation, differing only in the band-filling and gap size. Semiconductors, the basis of all modern electronics, are simply insulators with a small enough gap that thermal energy can promote a measurable number of electrons across it.

The Core Insight

In a periodic potential, electron energies form bands of allowed states separated by gaps of forbidden energies. At absolute zero, electrons fill these bands from the bottom up to the Fermi level. The classification follows:

Metal: the Fermi level lies inside a band (partially filled band) → many electrons can move into nearby empty states → conducts.
Insulator: the Fermi level lies in a large gap (≳ 4 eV); valence band full, conduction band empty → no available states for conduction → does not conduct.
Semiconductor: same as insulator but with a small gap (~0.1–3 eV) → thermal excitation populates the conduction band weakly → conducts moderately, strongly temperature-dependent.

The decisive quantities are the band gap $E_{g}$ and the band filling.

Formal Statement

Bloch's Theorem and Energy Bands

In a perfect crystal with a periodic potential $V (r + R) = V (r)$ for all lattice vectors $R$ , the stationary single-electron states have the Bloch form:

ψ_{n k} (r) = e^{i k \cdot r} u_{n k} (r), u_{n k} (r + R) = u_{n k} (r)

⟳Apply the lattice translation operator

T_{R}

, which commutes with

H

; simultaneous eigenstates satisfy

T_{R} ψ = e^{i k \cdot R} ψ

, forcing the plane-wave-times-periodic-function structure.

Here:

$k$ = crystal momentum (wavevector) in the first Brillouin zone,
$n$ = band index (labels distinct energy bands),
$u_{n k}$ = lattice-periodic envelope function,
$E_{n} (k)$ = dispersion relation giving the energy of band $n$ .

For each $n$ , as $k$ ranges over the Brillouin zone, $E_{n} (k)$ sweeps out a continuous band of allowed energies. Energy ranges with no solution for any $k$ are forbidden gaps.

Fermi Level and Classification

The Fermi–Dirac distribution gives the occupation probability of a state of energy $E$ at temperature $T$ :

f (E) = \frac{1}{exp ( \frac{E - μ}{k _{B} T} ) + 1}

⟳Maximize the entropy of a system of indistinguishable fermions subject to fixed total energy and particle number, using Lagrange multipliers

β = 1/ k_{B} T

and

μ

(the chemical potential ≈ Fermi level

E_{F}

T = 0

where $μ \to E_{F}$ as $T \to 0$ . At $T = 0$ , $f (E) = 1$ for $E < E_{F}$ and $0$ for $E > E_{F}$ . The classification is:

| Class | Band filling at $T = 0$ | Gap $E_{g}$ | |-------|----------------------|-----------| | Metal | partially filled band | (Fermi level inside band) | | Semiconductor | full valence, empty conduction | small (~0.1–3 eV) | | Insulator | full valence, empty conduction | large (≳ 4 eV) |

Band-Filling Counting Rule

A band derived from $N$ primitive cells holds exactly $2 N$ electron states (the factor 2 from spin). If the number of valence electrons per primitive cell is even, bands can be completely filled, yielding an insulator or semiconductor (if bands do not overlap). If odd, the topmost band is necessarily half-filled, guaranteeing metallic behavior.

Step 1. Periodic (Born–von Kármán) boundary conditions over

N

cells quantize

k

into exactly

N

allowed values within the first Brillouin zone. Step 2. Each allowed

k

contributes one orbital state per band; including the two spin projections gives

2 N

states per band. Step 3. Let

z

= valence electrons per primitive cell. The total electron count is

z N

. The number of fully filled bands needed is

z N / (2 N) = z /2

. Step 4. If

z

is even,

z /2

is an integer — bands fill exactly to a gap edge → insulator/semiconductor (absent band overlap). If

z

is odd,

z /2

is a half-integer — the topmost band is half-filled → metal.

■

Caveat (Step 4): band overlap can make an even- $z$ material a metal (e.g., divalent magnesium), so the rule predicts metals reliably but not insulators.

Felix Bloch | 1905–1983

Swiss-American physicist who, in his 1928 doctoral work under Heisenberg, proved that electrons in a periodic potential are described by plane waves modulated by a lattice-periodic function — now Bloch's theorem. This single result dissolved the paradox of why electrons in a metal travel almost freely despite a dense array of ion cores: a perfectly periodic lattice does not scatter Bloch waves at all; only deviations from periodicity (phonons, impurities) cause resistance. Bloch later won the 1952 Nobel Prize in Physics (with Purcell) for nuclear magnetic resonance, a separate achievement.

Building Intuition

Imagine bringing $N$ identical atoms together from infinity. Each isolated atom has a sharp energy level — say, the $3 s$ level. When the atoms are far apart, you have $N$ identical, degenerate $3 s$ levels. As you push them closer, the wavefunctions overlap; the Pauli principle forbids identical states, so the single level splits into $N$ closely-spaced sublevels. With $N \sim 1 0^{23}$ , these merge into a quasi-continuous band spanning a finite energy width.

Different atomic levels broaden into different bands. Between them lie gaps — energies no electron can have. Now fill the bands with the available electrons, lowest first, like water filling a series of buckets:

If the water line stops partway up a bucket (partial filling), tilting the system (applying a voltage) lets electrons slosh easily into nearby empty levels → metal.
If the water exactly fills a bucket to the brim and the next bucket is far above (big gap), tilting does nothing — there's nowhere to go → insulator.
If the gap to the next bucket is small, a little thermal "splashing" puts a few drops in the upper bucket → semiconductor.

The crucial picture: conduction requires empty states immediately above occupied ones. A full band carries no net current because for every electron moving one way, there's one moving the other.

Energy bands and the Fermi level for the three material classes

Worked Examples

Intrinsic carrier concentration in silicon

Given: Silicon at

T = 300

K with band gap

E_{g} = 1.12

eV. Effective densities of states

N_{c} = 2.8 \times 1 0^{19} cm^{- 3}

(conduction) and

N_{v} = 1.04 \times 1 0^{19} cm^{- 3}

(valence).

k_{B} T = 0.0259

eV. Find: the intrinsic carrier concentration

n_{i}

. Step 1. The intrinsic concentration is

n_{i} = N_{c} N_{v} exp (- \frac{E _{g}}{2 k _{B} T})

⟳In equilibrium

n p = N_{c} N_{v} exp (- E_{g} / k_{B} T)

(law of mass action); for an intrinsic semiconductor

n = p = n_{i}

, so take the square root.

Step 2. Compute the prefactor:

N_{c} N_{v} = (2.8 \times 1 0^{19}) (1.04 \times 1 0^{19}) = 2.91 \times 1 0^{38} = 1.71 \times 1 0^{19} cm^{- 3} .

Step 3. Compute the exponent:

\frac{E _{g}}{2 k _{B} T} = \frac{1.12}{2 ( 0.0259 )} = \frac{1.12}{0.0518} = 21.6, e^{- 21.6} = 4.1 \times 1 0^{- 10} .

Step 4. Multiply:

n_{i} = (1.71 \times 1 0^{19}) (4.1 \times 1 0^{- 10}) = 7.0 \times 1 0^{9} cm^{- 3} .

n_{i} \approx 7 \times 1 0^{9} cm^{- 3}

This is about 13 orders of magnitude below silicon's atomic density (~ $5 \times 1 0^{22} cm^{- 3}$ ) — explaining why pure silicon is a poor conductor and why doping (adding ~ $1 0^{16}$ impurities/cm³) so dramatically alters it.

Free-electron Fermi energy of sodium

Given: Sodium (monovalent,

z = 1

) with

n = 2.65 \times 1 0^{28} electrons/m^{3}

. Constants:

ℏ = 1.055 \times 1 0^{- 34}

J·s,

m_{e} = 9.11 \times 1 0^{- 31}

kg. Find: the Fermi energy

E_{F}

in eV. Step 1. The free-electron Fermi energy is

E_{F} = \frac{ℏ ^{2}}{2 m _{e}} (3 π^{2} n)^{2/3}

⟳Fill momentum states up to the Fermi sphere of radius

k_{F}

; require the number of states

2 \cdot V k_{F}^{3} / (6 π^{2}) = N

, giving

k_{F} = (3 π^{2} n)^{1/3}

, then

E_{F} = ℏ^{2} k_{F}^{2} /2 m_{e}

Step 2. Compute

k_{F}

3 π^{2} n = 3 (9.87) (2.65 \times 1 0^{28}) = 7.85 \times 1 0^{29} m^{- 3},

k_{F} = (7.85 \times 1 0^{29})^{1/3} = 9.22 \times 1 0^{9} m^{- 1} .

Step 3. Compute

E_{F}

E_{F} = \frac{( 1.055 \times 1 0 ^{- 34} ) ^{2}}{2 ( 9.11 \times 1 0 ^{- 31} )} (9.22 \times 1 0^{9})^{2} = \frac{1.113 \times 1 0 ^{- 68}}{1.822 \times 1 0 ^{- 30}} (8.50 \times 1 0^{19}) .

E_{F} = (6.11 \times 1 0^{- 39}) (8.50 \times 1 0^{19}) = 5.19 \times 1 0^{- 19} J .

Step 4. Convert:

5.19 \times 1 0^{- 19} /1.602 \times 1 0^{- 19} = 3.24

eV.

E_{F} \approx 3.2 eV

The corresponding Fermi temperature $T_{F} = E_{F} / k_{B} \approx 3.8 \times 1 0^{4}$ K — vastly larger than room temperature, which is why metals' conduction electrons behave as a degenerate quantum gas even at 300 K.

Common Misconceptions

"A full band can still conduct if you push hard enough." False. A completely filled band carries zero net current at any field strength (within band theory), because all states are occupied and Pauli blocking prevents the electron distribution from shifting in

k

-space. Current requires an asymmetric occupation, which demands accessible empty states. This is why insulators are insulators no matter the applied (sub-breakdown) voltage. "Semiconductors conduct better when cooled, like metals." Backwards. In a metal, resistivity rises with temperature (more phonon scattering). In an intrinsic semiconductor, conductivity rises exponentially with temperature because more carriers are thermally excited across the gap — the

n_{i} \propto e^{- E_{g} /2 k_{B} T}

factor dominates over scattering. This opposite temperature dependence is a key experimental fingerprint distinguishing the two classes. Even valence count does not guarantee an insulator. The band-filling rule predicts metals for odd

z

reliably, but for even

z

it predicts insulator/semiconductor only if bands do not overlap. Magnesium and zinc have

z = 2

(even), yet are excellent metals: their valence and conduction bands overlap in energy, so electrons spill from a "full" band into an overlapping empty one, leaving both partly filled. Thus band overlap — a feature of the actual 3D dispersion

E_{n} (k)

— can convert a predicted insulator into a (semi)metal.

Connections

Band theory sits at the crossroads of several frameworks. It builds directly on Bloch's Theorem, which guarantees the wave-like nature of electrons in a periodic lattice, and on the Free Electron Model, which it generalizes by adding the lattice potential. The shape of bands near gap edges is captured by the Nearly Free Electron Model and, complementarily, by the Tight Binding Model (which starts from atomic orbitals). Filling the bands uses Fermi-Dirac Statistics and the concept of the Density of States. The geometry of allowed $k$ lives in the Brillouin Zone, and the dynamics of carriers near band edges is governed by the Effective Mass tensor and the notion of Holes (Semiconductor Physics). For strongly interacting systems, band theory famously fails, motivating the study of Mott Insulators.

Real-World Applications

Silicon transistors & integrated circuits. Silicon's $1.12$ eV gap and its controllable doping (n-type/p-type) underpin the entire digital economy — every CPU, memory chip, and microcontroller relies on band-gap engineering.
Light-emitting diodes (LEDs). Direct-gap semiconductors (GaN $E_{g} \approx 3.4$ eV for blue, GaAs for infrared) emit photons of energy $\approx E_{g}$ when electrons recombine with holes across the gap.
Photovoltaic solar cells. A photon with energy $> E_{g}$ promotes an electron to the conduction band, creating an electron–hole pair that is separated by a p–n junction to drive current. Silicon's gap is near-optimal for the solar spectrum.
Photodetectors & infrared sensors. Narrow-gap materials (InSb, HgCdTe) absorb low-energy IR photons, used in thermal imaging and night-vision cameras.
Thermoelectric generators. Engineered band structures (e.g., Bi₂Te₃) with high carrier mobility but low thermal conductivity convert heat to electricity in spacecraft RTGs and waste-heat recovery.
Power electronics with wide-bandgap materials. SiC ( $E_{g} \approx 3.3$ eV) and GaN devices switch high voltages efficiently in electric-vehicle inverters and fast chargers, exploiting their large gaps for high breakdown fields.
Insulating dielectrics. Diamond ( $E_{g} \approx 5.5$ eV) and SiO₂ ( $\approx 9$ eV) serve as gate insulators and high-voltage barriers precisely because their gaps are too large to thermally populate.

Historical Development

1900 — Paul Drude introduces the classical free-electron (Drude) model of metals, explaining Ohm's law but failing on specific heat.
1926–27 — Arnold Sommerfeld quantizes the free-electron gas using Fermi–Dirac statistics, fixing the specific-heat anomaly but still ignoring the lattice.
1928 — Felix Bloch proves Bloch's theorem in his thesis, showing electrons in a periodic potential form propagating waves; the concept of energy bands is born.
1929 — Rudolf Peierls analyzes the origin of band gaps from Bragg reflection of electron waves and introduces the concept of holes (positive carriers).
1930 — Léon Brillouin formulates the Brillouin zone construction, clarifying the role of reciprocal-space boundaries in gap formation.
1931 — Alan Herries Wilson publishes the decisive papers distinguishing metals, semiconductors, and insulators by band filling and gap size — the founding statement of this article's topic.
1933 — Wigner & Seitz develop a practical cellular method for computing real band structures (sodium).
1947 — Bardeen, Brattain & Shockley invent the point-contact transistor at Bell Labs, the first triumphant technological application of semiconductor band theory (Nobel 1956).
1964–65 — Hohenberg, Kohn & Sham establish density functional theory, enabling accurate first-principles band-structure calculation for real materials.

Subtleties & Limits

Band theory is a single-electron approximation: it treats each electron as moving in an averaged, static periodic potential. This works astonishingly well for simple metals and semiconductors but fails when electron–electron interactions are strong. The flagship failure is the Mott insulator: materials like NiO have a half-filled $d$ -band and should be metals by the band-filling rule, yet they are robust insulators because strong on-site Coulomb repulsion freezes the electrons in place. Predicting which correlated materials are metals or insulators — and the related problems of high-temperature superconductivity, the band gaps of strongly correlated oxides, and the precise gap values from first principles (the notorious "band-gap problem" where standard DFT systematically underestimates gaps by ~50%) — remains an active frontier. Time-dependent and many-body methods (GW approximation, dynamical mean-field theory) extend band theory, but a fully predictive, computationally tractable theory of correlated electron materials is unsolved.