Electron-hole pair generation: Visual walkthrough — Hardware deep dive

Everything below rests on one question:

How many electrons manage to break free at temperature $T$ — and why does heat set that number?

Step 1 — Start at absolute zero: a frozen crystal

WHAT. We begin with a pure (intrinsic) silicon crystal at temperature $T = 0$ . Every silicon atom shares its outer electrons with neighbours in covalent bonds. Every electron is locked in a bond. Nothing is free to move.

WHY. To count free electrons we need a starting point where the count is exactly zero. Absolute zero gives us that clean baseline: no thermal energy, no escaped electrons, no holes. Every carrier we later find was generated — so counting generation = counting carriers.

PICTURE. Look at figure s01. On the left is the crystal: blue dots are atoms, the springy links between them are the shared-electron bonds — all intact. On the right is the energy-band picture of the same crystal: a lower shelf (the valence band, where bound electrons live) is completely full, an upper shelf (the conduction band, where free electrons roam) is completely empty, and between them is a forbidden gap of height $E_{g}$ — no electron is allowed to sit inside the gap.

Step 2 — Add heat: some electrons win the energy lottery

WHAT. Now warm the crystal to a real temperature (say room temperature, $300 K$ ). The atoms vibrate. These vibrations are little packets of energy sloshing around. Occasionally a bound electron catches a big enough packet — at least $E_{g}$ — and leaps across the gap. It is now free. Its empty seat becomes a hole.

WHY. We must translate "heat" into a number: what fraction of electrons get lucky? Heat is random, so we ask a probability question — "what is the chance a state high up is occupied?" — not a deterministic one.

PICTURE. In figure s02, a red arrow shows one electron jumping from the full lower shelf across the gap $E_{g}$ to the empty upper shelf. It leaves behind a coral circle — the hole. Crucially, one jump = one electron + one hole. They are born as a pair.

Step 3 — Why the probability is an exponential

WHAT. We need the chance that a state at energy $E$ above the mid-gap reference is occupied. For energies far above the typical packet size ( $E ≫ k T$ ), that chance falls off as $P (E) \propto e^{- E / k T} .$

WHY THIS TOOL — why $e^{- E / k T}$ and not, say, $1/ E$ ? Because thermal occupation follows the Fermi–Dirac distribution, and when a state sits many $k T$ above the Fermi level, that distribution flattens into the Boltzmann tail $e^{- E / k T}$ . The exponential is the unique function where each extra chunk of energy multiplies the difficulty by the same factor — climb one more $k T$ and your odds drop by the same ratio $e^{- 1} \approx 0.37$ every time. That "constant-factor-per-step" is precisely how random thermal energy stacks up.

PICTURE. Figure s03 plots $e^{- E / k T}$ against energy $E$ . Read the red curve left to right: at $E = 0$ the probability is 1 (certain), then it plummets. The dashed marker at $E = E_{g} /2$ shows the value we will actually need — and it is already tiny.

Step 4 — Where does the reference sit? The factor of 2 appears

WHAT. We must decide from where the electron measures its climb. In a pure semiconductor, the Fermi level $E_{F}$ — the energy where occupation probability is exactly $\frac{1}{2}$ — sits right in the middle of the gap. So a valence electron measured from $E_{F}$ only has to climb half the gap, $E_{g} /2$ , to reach the conduction-band edge.

WHY. The exponential of Step 3 measures energy from the Fermi level. If we naively wrote $E_{g}$ as the climb we would double-count: the reference is not the bottom of the gap, it is the middle. Getting this reference right is the single most common mistake in the whole topic.

PICTURE. Figure s04 stacks the band diagram with the Fermi level drawn as a dashed line dead-centre in the gap. A butter-coloured bracket marks the climb from $E_{F}$ up to the conduction edge — its height is labelled $E_{g} /2$ , not $E_{g}$ .

Step 5 — Not one state, but a whole shelf: $N_{c}$ and $N_{v}$

WHAT. A probability alone is not a count. To get a number per cubic centimetre, multiply the occupation probability by how many seats are available. The conduction band offers $N_{c}$ effective seats per unit volume; the valence band offers $N_{v}$ empty-seat slots per unit volume.

WHY. Probability answers "what fraction succeed?"; the density of states answers "out of how many?". Number = (how many tries) × (chance each succeeds). We cannot skip either.

PICTURE. Figure s05 shows the two shelves as rows of seats. The upper row has $N_{c}$ slots painted lavender (available conduction seats); the lower row has $N_{v}$ slots painted coral (valence seats that can become holes). Only a thin sliver of seats near each edge matters — that is what "effective density of states" bundles up.

Building the count for electrons ( $n$ ) and holes ( $p$ ) separately: $n \approx N_{c} e^{- (E_{c} - E_{F}) / k T}, p \approx N_{v} e^{- (E_{F} - E_{v}) / k T}$

$E_{c}$ — conduction-band edge energy; $E_{v}$ — valence-band edge energy.
$E_{c} - E_{F}$ — the electron's climb from Fermi level up; $E_{F} - E_{v}$ — the hole's climb from Fermi level down.
With $E_{F}$ mid-gap, both climbs equal $E_{g} /2$ .

Step 6 — Pairs, always pairs: $n = p$

WHAT. In a pure crystal, every electron that jumps up leaves exactly one hole behind. So the number of free electrons equals the number of holes: $n = p \equiv n_{i} .$ We give this common value its own name, $n_{i}$ , the intrinsic carrier concentration.

WHY. This is the accounting rule that lets us collapse two unknowns ( $n$ and $p$ ) into one ( $n_{i}$ ). Without purity this would fail — doping deliberately breaks $n = p$ — but for intrinsic material it holds exactly.

PICTURE. Figure s06 shows a tally: every red electron-dot on the upper shelf is paired by a coral hole-circle on the lower shelf, connected by a thin thread. Count the reds, count the corals — they match. That matching is $n = p = n_{i}$ .

Step 7 — Multiply, then take the root: the result appears

WHAT. Multiply the electron count and the hole count from Step 5: $n p = N_{c} N_{v} e^{- (E_{c} - E_{F}) / k T} e^{- (E_{F} - E_{v}) / k T} = N_{c} N_{v} e^{- (E_{c} - E_{v}) / k T} .$ Notice $E_{c} - E_{v} = E_{g}$ — the full gap! The Fermi level $E_{F}$ cancels out completely. So $n p = N_{c} N_{v} e^{- E_{g} / k T} .$ Because $n = p = n_{i}$ (Step 6), the left side is $n_{i}^{2}$ : $n_{i}^{2} = N_{c} N_{v} e^{- E_{g} / k T} ⟹ n_{i} = N_{c} N_{v} e^{- E_{g} /2 k T} .$

WHY THE PRODUCT. Multiplying $n$ by $p$ makes the unknown Fermi level vanish — a beautiful trick. The product only cares about the whole gap $E_{g}$ , no reference needed. This is the mass-action law $n p = n_{i}^{2}$ .

WHY THE SQUARE ROOT. The product carries the full gap ( $e^{- E_{g} / k T}$ ). But $n_{i}$ is one carrier type, which we get by taking $n_{i}^{2}$ . The square root of $e^{- E_{g} / k T}$ is $e^{- E_{g} /2 k T}$ — and that is the factor of 2 from Step 4, arriving a second way. Two independent routes, same 2: the physics is consistent.

PICTURE. Figure s07 draws the algebra as a flow: two exponentials merge (Fermi level cancels), giving $n p$ ; then a square-root gate splits it back to $n_{i}$ , dropping the exponent to $E_{g} /2 k T$ .

Step 8 — Degenerate & limiting cases (the corners the formula must survive)

WHAT & WHY. A formula you trust must behave sensibly at its extremes. We check the corners so you never hit an unexplained scenario.

PICTURE. Figure s08 plots $n_{i}$ versus temperature $T$ on a log axis, marking the three corners below.

$T \to 0$ (absolute zero). The exponent $- E_{g} /2 k T \to - \infty$ , so $e^{- E_{g} /2 k T} \to 0$ and $n_{i} \to 0$ . The crystal is a perfect insulator — matches Step 1 exactly. Good.
$T \to \infty$ (very hot). The exponent $\to 0$ , so $e^{0} = 1$ and $n_{i} \to N_{c} N_{v}$ : carriers saturate at the number of available seats. You cannot free more electrons than there are seats — physically sane.
$E_{g} \to 0$ (a metal-like closed gap). The exponential $\to 1$ at any $T$ ; carriers are abundant even when cold. This is why gapless/tiny-gap materials conduct like metals — no barrier to climb.
Large $E_{g}$ (insulator, e.g. diamond $\approx 5.5 eV$ ). The exponent is enormous and negative → $n_{i}$ is astronomically tiny → essentially no free carriers → a genuine insulator. The same formula smoothly spans semiconductors and insulators just by changing $E_{g}$ .

The one-picture summary

Figure s09 compresses the entire walkthrough into a single frame: the cold full-band start (Step 1), the thermal jump making a pair (Steps 2–3), the mid-gap Fermi reference giving the half-gap climb (Step 4), the seat-counting prefactor (Step 5), pair equality (Step 6), and the product-then-root that lands on the boxed result (Step 7) — with the $T \to 0$ and $T \to \infty$ corners flagged (Step 8).

Recall Feynman retelling: the whole derivation in plain words

Start with a frozen crystal — every electron glued into a bond, zero free carriers. Now heat it: random vibrations occasionally hand an electron enough energy ( $E_{g}$ ) to leap the forbidden gap. Each leap makes a free electron and an empty seat (a hole) — a pair. How many leaps? Heat is random, so the chance of a big enough leap dies off exponentially — every extra chunk of energy multiplies the difficulty by the same factor, giving $e^{- cost / k T}$ . But the electron measures its climb from the Fermi level, which sits smack in the middle of the gap, so it climbs only half the gap: the exponent is $E_{g} /2 k T$ . To turn a chance into a count, multiply by how many seats exist ( $N_{c}$ up top, $N_{v}$ down below). Because the crystal is pure, electrons and holes come in equal numbers, both called $n_{i}$ . The clever move: multiply the electron count by the hole count — the unknown Fermi level cancels, leaving the full gap $E_{g}$ ; then take the square root to get back one carrier type. The root turns $e^{- E_{g} / k T}$ into $e^{- E_{g} /2 k T}$ — the same factor of 2, confirmed twice. Final answer: $n_{i} = N_{c} N_{v} e^{- E_{g} /2 k T}$ . Check the corners: cold → zero carriers (insulator), hot → carriers saturate at the seat count. It all fits.

Visual walkthrough — Electron-hole pair generation