Intrinsic vs extrinsic semiconductors: Visual walkthrough

We only assume you can count, you know that heat = jiggling, and you know a battery pushes charge. Everything else — including what $n_{i}$ and $E_{g}$ actually mean, and the constants that turn temperature into energy — is built below before it is used.

Step 1 — The frozen crystal: why "no carriers" at $T = 0$

WHAT. Picture silicon at absolute zero (the coldest possible, $T = 0$ K). Every silicon atom shares its 4 outer electrons with 4 neighbours. Each shared pair is a bond — think of it as two hands clasped together. Every hand is busy.

WHY start here. If we want to count free charges later, we first need a state with zero free charges as our baseline. That baseline is the frozen crystal.

PICTURE. Below, every electron (small teal dot) sits on a bond line. Not one is loose. A voltage arrow pushes — and nothing flows, because a clasped hand can't reach for the battery.

Figure — Intrinsic vs extrinsic semiconductors

Step 2 — Heat breaks a bond: one kick makes two carriers

WHAT. Warm the crystal ( $T > 0$ ). Atoms vibrate. Every so often a vibration is violent enough to rip an electron out of its bond and fling it up to the high shelf.

WHY this matters. This single event is the only source of carriers in a pure crystal, and it always produces two things at once. That "two-ness" is the seed of the whole formula.

PICTURE. Follow the orange arrow: one electron leaves its bond (arrives on the conduction shelf, now free to drift). It leaves behind an empty clasp — a hole (plum circle) that acts like a loose $+$ charge.

Step 3 — From Fermi–Dirac to a simple exponential, and why it's allowed

WHAT. Every place an electron is allowed to sit is called a quantum state (a "seat"), and each seat has a definite energy. We write == $E$ for the energy of one such seat (measured in electron-volts, eV). We also need one landmark energy: the Fermi energy $E_{F}$ == — the "sea level" of the electron ocean, the energy at which a seat has a 50-50 chance of being occupied. Higher-than- $E_{F}$ seats are usually empty; lower ones are usually full.

Now: how likely is one particular seat, at energy $E$ , to actually hold an electron? The exact rule (see Fermi level and Fermi-Dirac statistics) is the Fermi–Dirac occupancy: $f (E) = \frac{1}{1 + e ^{(E - E_{F}) / k T}}$ Two brand-new letters appear in that exponent, so we name them before leaning on them:

== $k$ is Boltzmann's constant==, $k \approx 8.62 \times 1 0^{- 5} eV/K$ — the fixed exchange rate that converts a temperature into an energy.
== $T$ is the absolute temperature in kelvin. Their product $k T$ is the typical thermal energy of one random "knock"== ( $\approx 0.026$ eV at room temperature).

When the seat sits far above sea level — meaning $E - E_{F} ≫ k T$ — the exponential term is huge, the $+ 1$ is negligible, and this collapses to the Boltzmann factor: $f (E) \approx e^{- (E - E_{F}) / k T}$

WHY this simplification is legal here. In silicon the shelf is a full gap away from sea level, and one gap is many thermal knocks: $E_{g} \approx 1.12 eV$ while $k T \approx 0.026 eV$ at room temperature, so $E_{g} / k T \approx 43 ≫ 1$ . That is exactly the "far above sea level" condition, so Fermi–Dirac legitimately reduces to the simple exponential. (If a shelf ever sat within a few $k T$ of $E_{F}$ — heavy doping, Step 8 — this shortcut breaks and we'd keep the full $f (E)$ .)

WHY an exponential at all, not something gentler. Energy comes in random thermal knocks. To collect a big energy you need many lucky knocks in a row — and "many lucky things in a row" always multiplies into an exponential drop. That is exactly the behaviour we see (diamond insulates, germanium leaks), so the exponential is the right tool.

PICTURE. The teal curve is the true Fermi–Dirac $f (E)$ ; the plum dashed curve is the Boltzmann approximation. See how they lie on top of each other once you are more than a few $k T$ above $E_{F}$ .

Step 4 — Counting electrons on each shelf: where $N_{c}, N_{v}$ come from

WHAT. Number on a shelf = (how many seats that shelf effectively offers) × (chance a seat is filled). Using $E_{c}$ for the bottom of the high shelf, $E_{v}$ for the top of the low shelf, and the Boltzmann factor from Step 3 (evaluated at the shelf edge, since carriers pile up near the edge):

$n = N_{c} e^{- (E_{c} - E_{F}) / k T}, p = N_{v} e^{- (E_{F} - E_{v}) / k T}$

Both formulas inherit Step 3's licence: they need $E_{c} - E_{F} ≫ k T$ and $E_{F} - E_{v} ≫ k T$ — true for a lightly/moderately doped or pure crystal, where $E_{F}$ sits near mid-gap.

WHY $N_{c}$ and $N_{v}$ are not just "the number of atoms." A shelf is not a single energy — it is a whole spread of seats stacking upward, and quantum mechanics tells us how densely seats are packed at each energy (the density of states). Adding up "seats × their Boltzmann weight" across that spread is an integral, and the whole integral tidies into one number times the edge exponential. That number is the effective density of states: $N_{c} = 2 (\frac{2 π m _{e}^{*} k T}{h ^{2}})^{3/2}, N_{v} = 2 (\frac{2 π m _{h}^{*} k T}{h ^{2}})^{3/2}$ Three symbols enter here, each named before use:

== $h$ is Planck's constant==, $h \approx 6.63 \times 1 0^{- 34} J\cdots$ — the tiny fixed number that sets the "grain size" of the quantum world and thus how finely seats can be packed.
== $m_{e}^{*}$ is the electron effective mass and $m_{h}^{*}$ is the hole effective mass== — how "heavy" each carrier behaves once you account for the crystal's internal forces (not the free-space mass).

WHY it scales as $T^{3/2}$ and uses an effective mass. The effective mass $m^{*}$ captures how "heavy" the carrier behaves inside the crystal's forces — heavier carriers pack seats more densely. The $T^{3/2}$ appears because a hotter crystal spreads carriers over a taller slice of the shelf, opening more usable seats. So $N_{c}, N_{v}$ are "effective seat counts" that themselves grow slowly with temperature.

PICTURE. An energy ladder: $E_{v}$ at bottom, $E_{c}$ at top, gap $E_{g} = E_{c} - E_{v}$ between them, and $E_{F}$ near the middle. The shaded fans show the density of states widening away from each edge; the two arrows show the two distances the exponentials measure.

Step 5 — The multiply trick: killing $E_{F}$

WHAT. Multiply the two lines from Step 4 together: $n p = N_{c} N_{v} e^{- (E_{c} - E_{F}) / k T} e^{- (E_{F} - E_{v}) / k T}$

WHY multiply (and not add or subtract). Look at the exponents: one carries $+ E_{F}$ , the other carries $- E_{F}$ . When you multiply exponentials you add their exponents — and $(- (E_{c} - E_{F})) + (- (E_{F} - E_{v}))$ cancels $E_{F}$ completely, leaving only $- (E_{c} - E_{v}) = - E_{g}$ . We didn't know $E_{F}$ , so the smartest move is to build a quantity that doesn't care about it. Multiplication is the tool that does that.

PICTURE. Watch the two $E_{F}$ arrows below — one points up, one points down; laid tip-to-tail they exactly cancel, and only the full gap $E_{g}$ survives.

$n p = N_{c} N_{v} e^{- (E_{c} - E_{v}) / k T} = N_{c} N_{v} e^{- E_{g} / k T}$

The two exponents added → the $E_{F}$ terms vanished.
$E_{c} - E_{v}$ collapsed into the single gap $E_{g}$ .
What remains depends only on things we can look up: $N_{c}, N_{v}, E_{g}, T$ .

Step 6 — Closing the loop for a pure crystal

WHAT. Now bring back the pure-crystal fact from Step 2: $n = p$ . Call that common value $n_{i}$ . Then $n p = n_{i} \cdot n_{i} = n_{i}^{2}$ , so: $n_{i}^{2} = N_{c} N_{v} e^{- E_{g} / k T} ⟹ n_{i} = N_{c} N_{v} e^{- E_{g} /2 k T}$

WHY the square root and the 2. Since $n p = n_{i}^{2}$ and $n = p$ , taking the square root of both sides splits the exponent in half: $e^{- E_{g} / k T}$ becomes $e^{- E_{g} /2 k T}$ . The 2 is simply "one gap shared between one electron and one hole."

PICTURE. A number-line comparison: the same crystal at three temperatures, with the height of the bar showing $n_{i}$ growing steeply as $T$ rises.

Step 7 — The degenerate & edge cases (so nothing surprises you)

WHAT. Push the formula to its extremes and check it still tells the truth.

WHY. A formula you trust only in the middle isn't trusted at all. We test the corners.

Case	Put into $n_{i} = N_{c} N_{v} e^{- E_{g} /2 k T}$	Result	Makes sense?
$T \to 0$ (frozen, Step 1)	exponent $\to - \infty$	$n_{i} \to 0$	✅ no carriers, matches the frozen picture
$E_{g}$ huge (diamond)	exponent hugely negative	$n_{i} \to 0$	✅ insulator
$E_{g}$ tiny (germanium)	exponent near 0	$n_{i}$ large	✅ leaky
$T$ very large	exponent $\to 0$	$n_{i} \to N_{c} N_{v}$	✅ saturates at the seat count

PICTURE. The $n_{i}$ -vs- $T$ curve with both limiting tails labelled: it hugs zero when cold and bends up steeply when hot, exactly as the exponent demands. This is the root of Temperature dependence of resistance in semiconductors.

Step 8 — When the whole formula stops being valid

WHAT. Every result above rests on two hidden hinges. Name them, so you know exactly where the ground ends.

WHY. A physicist's formula always comes with a "warranty." Using it outside the warranty gives confident-looking nonsense.

Heavy doping (degenerate case). If we dope so hard that $E_{F}$ climbs into the conduction shelf (or drops into the valence shelf), then $E_{c} - E_{F}$ is no longer $≫ k T$ , Step 3's shortcut fails, and $n = N_{c} e^{- (E_{c} - E_{F}) / k T}$ over-counts. Real heavily-doped material needs the full Fermi–Dirac integral. The Law of Mass Action $n p = n_{i}^{2}$ itself also drifts (band-gap narrowing).
Vanishing or negative gap ( $E_{g} \to 0$ ). If a material has $E_{g} = 0$ (a semimetal) or the bands overlap, there is no "climb" left: carriers exist even at $T = 0$ , $n_{i}$ no longer describes the situation, and $e^{- E_{g} /2 k T}$ loses its meaning. Our whole picture assumed a real, positive gap that is $≫ k T$ .
Not at equilibrium. Shine light on it, or push current through it, and $n p$ can exceed $n_{i}^{2}$ . The mass-action law is an equilibrium statement only — the launching point for PN junction diode and Carrier mobility and drift current.

PICTURE. A validity map: gap on the horizontal axis, doping (how close $E_{F}$ is to a band) on the vertical. The safe "warm-paper" island in the middle is where every formula on this page holds.

The one-picture summary

Everything above in a single frame: a bond breaks (Step 2) → Fermi–Dirac simplifies to $e^{- (E - E_{F}) / k T}$ because the gap $≫ k T$ (Step 3) → two shelf-counts built from the density of states $N_{c}, N_{v}$ (Step 4) → multiply to kill $E_{F}$ (Step 5) → set $n = p$ to get $n_{i}$ (Step 6) → valid only inside the warranty (Step 8). This same box, applied to a doped crystal (where $n \neq = p$ ), is what lets you compute minority carriers and ultimately the PN junction diode.

Recall Feynman retelling — the whole walkthrough in plain words

Start with a crystal so cold that every kid in the room is holding both hands with a neighbour. Push with a battery — nobody budges, because busy hands can't reach it. Now warm the room. Once in a while a kid gets jostled hard enough to let go and wander off (a free electron), leaving an empty hand-clasp that other kids keep filling and re-emptying, so the empty spot itself seems to stroll around (a hole). One jostle always makes exactly one wanderer and one empty spot — so wanderers and empty spots come in equal numbers. How often does a jostle succeed? The exact chance is the Fermi–Dirac rule, but because the wall (the gap) is way taller than a typical jostle (that "typical jostle" size is $k T$ , temperature turned into energy by Boltzmann's constant $k$ ), that fancy rule simplifies to the plain $e^{- climb / k T}$ curve — small for a tall wall, big for a short one. Counting wanderers up top and empty spots down below needs "how many seats each shelf offers," and those seat-counts $N_{c}, N_{v}$ come from stacking up quantum states (Planck's constant $h$ sets how finely they pack) — they even creep up as $T^{3/2}$ . Both counts are haunted by the unknown "sea level" $E_{F}$ . The trick: multiply them, and the two $E_{F}$ 's (one up, one down) cancel, leaving a clean law that survives even after we start doping. Finally, for the pure crystal we use "equal numbers" to trade $n p$ for $n_{i}^{2}$ , take a square root, and out pops $n_{i} = N_{c} N_{v} e^{- E_{g} /2 k T}$ — zero when frozen, soaring when hot. Just remember the warranty: it holds only while the shelves stay far from sea level and the gap is real and positive.