Defects — Schottky, Frenkel; non-stoichiometric defects: Visual walkthrough

Step 1 — The perfect crystal at absolute zero

WHAT. Picture a tidy grid of ions. Every site is filled. Nothing is out of place. This is a crystal at == $0 K$ == (absolute zero — the coldest possible, where thermal jiggling stops). See Crystal Lattices and Unit Cells for the grid itself.

WHY start here. Because a perfect crystal is the one arrangement we can count exactly: there is only one way to have everything perfect. That "one way" is the seed of the whole argument — disorder is measured relative to it.

PICTURE. The blue grid below. Every white dot is an ion on its home site. Count the ways to arrange it: exactly $1$ .

Figure — Defects — Schottky, Frenkel; non-stoichiometric defects

Step 2 — Making one vacancy costs energy

WHAT. Lift one ion out of the grid and set it aside. The empty spot it leaves is a vacancy. Doing this costs a fixed amount of energy, which we call $E$ per vacancy.

WHY it costs energy. The ion was held by its neighbours (electrostatic attraction). Pulling it out breaks those bonds — you must pay energy, exactly like lifting a fridge magnet off the door.

PICTURE. One amber dot has been removed; the energy meter on the right shows the cost climbing by $E$ for each removal.

Step 3 — But disorder pays you back: counting the arrangements

WHAT. Ask: in how many different ways can I scatter $n$ empty spots among $N$ sites? Call that number $W$ (for "ways"). One vacancy → $N$ choices. Two vacancies → far more. Many vacancies → an astronomically large $W$ .

WHY this matters. Nature does not just count energy; it counts possibilities. A state you can reach in a trillion ways is overwhelmingly more likely than one reachable in a single way. More arrangements = more entropy. (See Entropy and Gibbs Free Energy.)

PICTURE. Three ways to place one vacancy in a small grid — nature has no reason to prefer one, and there are many.

Step 4 — Turn the count into entropy with a logarithm

WHAT. Entropy is defined as $S = k_{B} ln W$ .

WHY a logarithm — and why this tool? Two independent crystals side by side have $W_{total} = W_{1} \times W_{2}$ (multiply the choices). But their disorder should add, not multiply. What turns multiplication into addition? The logarithm: $ln (W_{1} W_{2}) = ln W_{1} + ln W_{2}$ . That single property is the reason $ln$ appears — no other function makes "combine two systems" into "add their entropies."

PICTURE. Left: $W$ explodes upward (billions). Right: $ln W$ tames it into a gentle, addable curve.

Step 5 — Simplify $ln W$ with Stirling's approximation

WHAT. The factorials in $W$ are unusable directly ( $1 0^{23}!$ has no meaning on a calculator). Stirling's approximation replaces them: $ln x! \approx x ln x - x (for large x) .$

WHY. We need $ln W$ as a smooth formula we can do calculus on. Stirling turns clunky factorials into ordinary $x ln x$ terms — differentiable and clean. It is accurate precisely because $N$ and $n$ are enormous.

PICTURE. The jagged staircase of $ln x!$ hugged by the smooth Stirling curve $x ln x - x$ — nearly identical once $x$ is large.

Applying Stirling to each of the three factorials in $ln W = ln N! - ln n! - ln (N - n)!$ :

$ln W \approx [N ln N - N] - [n ln n - n] - [(N - n) ln (N - n) - (N - n)]$

Now watch the leftover linear (the " $- x$ ") pieces. Gather them separately: $from N! - N from n! + n from (N - n)! + (N - n) = - N + n + N - n = 0.$ The $+ n$ and $- n$ kill each other, and the $- N$ and $+ N$ kill each other — the linear terms exactly cancel to zero. Only the $x ln x$ pieces survive:

$ln W \approx N ln N - n ln n - (N - n) ln (N - n)$

Step 6 — Minimise Gibbs free energy: the tug-of-war

WHAT. Nature settles at the $n$ that makes Gibbs free energy smallest: $G = H - T S = n E - k_{B} T [N ln N - n ln n - (N - n) ln (N - n)] .$

WHY minimise $G$ (and why $G$ , not $U$ or Helmholtz $A$ ). Our crystal is held at constant temperature and pressure (a lab bench). Under exactly those two constraints, the state nature relaxes to is the one with the lowest Gibbs free energy $G = H - T S$ — this is the master rule of chemical thermodynamics (Entropy and Gibbs Free Energy). (At constant volume we would minimise Helmholtz $A = U - T S$ instead — different constraint, different function. We use $G$ because pressure, not volume, is what is fixed.) The term $n E$ pushes $n$ down (defects cost enthalpy); the $- T S$ term pushes $n$ up (defects add disorder, and $T$ amplifies the reward). The winner is a compromise.

PICTURE. Three curves versus $n$ : energy $H$ rising (amber), $- T S$ diving (cyan), and their sum $G$ dipping to a clear minimum (white) at some best value $n_{eq}$ .

Step 7 — Do the calculus: set the slope to zero

WHAT. The minimum of $G$ is where its slope with respect to $n$ is zero. The slope is the derivative $\frac{d G}{d n}$ .

WHY a derivative? A derivative answers "how fast does $G$ change if I add one more defect?" At the very bottom of the dip, adding a defect changes $G$ by essentially nothing — the slope is flat, i.e. zero. That flat point is the answer we want. No other tool locates the bottom of a curve.

PICTURE. Zoom on the minimum: the tangent line is perfectly horizontal — slope $= 0$ .

Differentiate term by term. Using $\frac{d}{d n} (n ln n) = ln n + 1$ :

$\frac{d G}{d n} = E - k_{B} T [- (ln n + 1) + (ln (N - n) + 1)] = E - k_{B} T ln \frac{N - n}{n}$

Set it to $0$ :

$E = k_{B} T ln \frac{N - n}{n}$

$E$ :: the enthalpy price of one vacancy.
$ln \frac{N - n}{n}$ :: the "disorder push-back" — big when $n$ is small (lots of room to disorder), small when $n$ is large.
At balance these two are equal — the tug-of-war is a draw.

Step 8 — Solve for $n$ : the exponential appears

WHAT. Undo the logarithm. The inverse of $ln$ is the exponential $e^{(\cdot)}$ . Applying it: $\frac{N - n}{n} = e^{E / k_{B} T} .$ Since defects are rare, $n ≪ N$ , so $N - n \approx N$ :

$n \approx N e^{- E / k_{B} T}$

WHY the exponential. Because $e^{(\cdot)}$ is exactly the function that cancels $ln (\cdot)$ — it answers "what number has this logarithm?" That is the only way to free $n$ from inside the log.

PICTURE. $n$ versus $T$ : dead-flat zero at low $T$ , then a sudden explosive climb — the signature exponential rocket.

Step 9 — From one vacancy to the boxed Schottky/Frenkel forms

WHAT. A Schottky defect is not one vacancy but a pair: one cation vacancy and one anion vacancy, created together to keep the crystal neutral. So we must count two separate vacancy sub-systems and then combine them.

WHY the counts multiply. Picture placing the $n$ cation vacancies (that's $W_{c}$ ways, from Step 3 with $N_{c}$ ) and, independently, the $n$ anion vacancies ( $W_{a}$ ways, from $N_{a}$ ). Every cation arrangement can pair with every anion arrangement, so the total ways multiply: $W = W_{c} \times W_{a}$ — exactly the "independent systems multiply" rule from Step 4.

WHY the energy halves. Redo Steps 4–8 with $W = W_{c} W_{a}$ . The logarithm splits the product into a sum $ln W_{c} + ln W_{a}$ , so the entropy is the sum of two identical Step-8 problems. Each half carries half of the pair energy $E_{s}$ — i.e. $E_{s} /2$ per vacancy. Solving each half gives an $e^{- E_{s} /2 k_{B} T}$ , and multiplying the two site-counts under one $n$ produces the geometric mean $N_{c} N_{a}$ :

$n \approx N_{c} N_{a} e^{- E_{s} /2 k_{B} T}$

$N_{c} N_{a}$ :: the combined ceiling from two independent site-lists; collapses to $N$ when $N_{c} = N_{a} = N$ .
$E_{s} /2$ :: the pair price $E_{s}$ shared equally between its two vacancies.

Frenkel is the same machine with one lattice vacancy ( $N$ sites) and one interstitial ion ( $N_{i}$ holes): $n \approx N N_{i} e^{- E_{f} /2 k_{B} T} .$

PICTURE. Side-by-side: Schottky's two removed sites (cation + anion) versus Frenkel's one moved ion (vacancy + interstitial).

The one-picture summary

Everything on one canvas: cost climbs (amber), disorder rewards (cyan), their fight settles $G$ at a minimum, and out drops the exponential defect count $n \approx N e^{- E / k_{B} T}$ that starts at zero when cold and rockets up when hot — then Step 9 doubles it up into the Schottky/Frenkel forms.

Recall Feynman: the whole walkthrough in plain words

Start with a perfect grid — boring, only one way to arrange it, and it's the cheapest. Now lift one ion out: you paid an energy $E$ (an enthalpy, because the lab is at fixed pressure), but suddenly there are $N$ places that empty spot could be — a windfall of possibilities. Nature loves possibilities, and heat makes it love them more. So it keeps making defects... but each new one buys less fresh disorder while charging the same price $E$ . There's a point where "not worth it" kicks in — that's the bottom of the free-energy valley, found by asking "where is the ground flat?" (slope zero) and then checking the ground curves upward on both sides (a real bottom). Solving that balance, the logarithm from counting flips into an exponential, and you get $n \approx N e^{- E / k_{B} T}$ : almost no defects when it's cold, a flood of them when it's hot. Schottky just does this twice (cation and anion), which — because independent counts multiply and their logs add — is where the square root and the $E_{s} /2$ come from.

Recall Quick checks

At $T = 0$ , how many defects? ::: Zero — the exponent $- E / k_{B} T \to - \infty$ , so $e^{- \infty} = 0$ . Perfect crystal. Why is the vacancy cost an enthalpy not an internal energy? ::: The crystal is at constant pressure, so any pressure–volume work is folded in; the constant-pressure energy is enthalpy $H$ , matching the rule "minimise $G = H - T S$ ". Why does the logarithm appear in entropy? ::: Because disorder must add when systems combine, and $ln$ turns the multiplying count $W_{1} W_{2}$ into $ln W_{1} + ln W_{2}$ . Why does the exponential appear at the end? ::: It is the inverse of $ln$ ; applying $e^{(\cdot)}$ frees $n$ from inside the logarithm. How do we know the flat point is a minimum? ::: The second derivative $\frac{d ^{2} G}{d n ^{2}} = k_{B} T (\frac{1}{n} + \frac{1}{N - n}) > 0$ , so the curve is bowl-shaped (convex) between $n = 0$ and $n = N$ . Where does the $/2$ in $e^{- E_{s} /2 k_{B} T}$ come from? ::: Splitting the pair-formation enthalpy $E_{s}$ between the two independent vacancies (cation + anion).