When the Impurity Takes Over
In guide 2 you watched a pure crystal make its own disorder from nothing but heat — the Schottky and Frenkel defects, the intrinsic kind. But there was a quiet warning buried in the numbers: in MgO the energy to make an intrinsic pair is near 6 eV, so at an ordinary firing or service temperature the crystal makes almost none of them. That vacuum gets filled by something else. Even a few hundred parts per million of an impurity — and every real powder has that — creates far more defects than heat alone ever could. These are the extrinsic defects, and at the temperatures where ceramics actually work, they, not the intrinsic ones, run the show.
If an impurity is going to dominate anyway, an engineer might as well choose it. That is aliovalent doping: deliberately dissolving an ion of a different valence into a host, so the crystal is forced to make compensating defects to stay neutral. Compare it with isovalent doping — swap Sr2+ for Ba2+ in BaTiO3 and nothing has to change, because +2 replaced +2 and the charge books never wobbled. Aliovalent doping is the opposite on purpose: put a +3 where a +4 belonged, or a +2 where a +4 belonged, and you leave the crystal one or two positive charges short. Something must appear to pay the debt, and what appears is the whole game.
Acceptors, Donors, and Who Pays the Charge
Aliovalent dopants come in two flavours. An acceptor is a lower-valence cation on the host site — a Ca2+ or Y3+ sitting where a Zr4+ belonged — leaving the site short of positive charge. The crystal can settle that debt with an oxygen vacancy: CaO -> Ca(Zr)'' + V(O) + O(O)x, or Y2O3 -> 2 Y(Zr)' + V(O) + 3 O(O)x. One Ca2+ buys one whole oxygen vacancy; two Y3+ share one. Those engineered vacancies are the seats an O2- hops between, so acceptor-doped zirconia becomes yttria-stabilized zirconia, the oxide-ion conductor inside fuel cells and oxygen sensors — around 8 mol% Y2O3 gives roughly 0.1 S/cm near 1000 degrees C.
A donor is the mirror image: a higher-valence cation on the host site — a Nb5+ where a Ti4+ belonged, or a La3+ where a Ba2+ belonged — bringing an extra positive charge. Now the debt is a surplus of positive, and the crystal can pay it two ways. It can make a cation vacancy (ionic compensation), or it can simply hand the extra charge to an electron (electronic compensation): Nb2O5 -> 2 Nb(Ti)* + 2 e' + 4 O(O)x + 1/2 O2, which turns TiO2 into an n-type semiconductor. Doping BaTiO3 with a little La3+ does the same, making it lightly conducting — the trick behind a charge-compensated PTC thermistor and semiconducting capacitor dielectrics.
- Compare valences. Is the dopant cation LOWER than the host it replaces (an acceptor) or HIGHER (a donor)? Ca2+ into Zr4+ is an acceptor; Nb5+ into Ti4+ is a donor.
- Write the dopant on its site in Kroger-Vink with its effective charge: Ca2+ on Zr4+ is short two positives, so Ca(Zr)''; Nb5+ on Ti4+ has one extra, so Nb(Ti)*.
- Offer a compensating defect of the OPPOSITE sign. For an acceptor: an oxygen vacancy V(O)** (ionic) or a hole h* (electronic). For a donor: a cation vacancy (ionic) or an electron e' (electronic).
- Decide WHICH compensation wins — this is not free choice but set by temperature and oxygen pressure. Acceptors lean ionic in low oxygen, electronic (holes) in high oxygen; donors lean electronic (n-type) unless the oxygen pressure is high enough to force cation vacancies.
- Finish with the guide-3 checklist: balance mass, balance sites (keep the cation-to-anion site ratio), and balance charge. Only then is the doping reaction real.
Notice the phrase that keeps returning: ionic or electronic, decided by temperature and gas. That branch is the hinge of this whole guide. A dopant fixes how many oxygen vacancies exist — one per Ca2+ is not a suggestion, it is arithmetic — but whether those vacancies stay put as ion carriers or hand their charge to electrons and holes depends on the atmosphere the crystal breathes. To see how the furnace gets a vote, we have to let the oxygen itself become a reactant.
A Formula That Breathes
Many oxides do not hold their tidy formula at all — they gain or lose oxygen to match the gas around them, a drift called nonstoichiometry. Iron oxide is the classic offender: it is never exactly FeO but always metal-deficient Fe(1-x)O, with x running from about 0.05 to 0.15. Titanium dioxide heated in a low-oxygen furnace becomes TiO2-x, quietly shedding oxygen and darkening from white to blue-black. Nickel oxide baked in air becomes nickel-deficient Ni(1-x)O and turns from green toward black. In each case the crystal has simply exchanged atoms with the atmosphere until its composition matched the oxygen pressure it was sitting in.
Written in the guide-3 grammar, these are just two reactions run forward or backward. Losing oxygen (reduction) is O(O)x -> V(O)** + 2 e' + 1/2 O2: an oxygen leaves as gas and abandons two electrons, so TiO2-x grows oxygen vacancies AND free electrons — an n-type semiconductor. Gaining oxygen (oxidation) is 1/2 O2 -> O(O)x + V(Fe)'' + 2 h*: oxygen from the gas builds a new lattice site, which demands a cation vacancy for site balance and leaves two holes behind — Fe(1-x)O and Ni(1-x)O grow cation vacancies AND holes, a p-type semiconductor. The same crystal is n-type in a reducing furnace and p-type in an oxidizing one; the atmosphere alone flips its electrical sign.
Mass Action and the Brouwer Map
How do we turn 'depends on the atmosphere' into a number? By treating every defect reaction as an ordinary chemical equilibrium and applying the law of mass action. The reduction reaction gets an equilibrium constant K = [V(O)**] times n^2 times (pO2)^(1/2) = exp(-delta-G/(k times T)), where n is the electron concentration and pO2 is the oxygen partial pressure. Because pO2 appears right inside the equilibrium, the crystal's defect count really is a function of the gas. Solve every such equilibrium at once, together with the charge-neutrality condition, and you know exactly how many of each defect the crystal holds at a given temperature and gas.
BROUWER MAP of an acceptor-doped oxide (log[defect] vs log pO2) low pO2 <---------------- oxygen pressure ----------------> high pO2 (reducing furnace) (oxidizing air) | n-TYPE REGION | DOPANT-FIXED PLATEAU | p-TYPE REGION | |------------------|--------------------------|------------------| | e' and V(O) | [V(O)] = const, | h* climb as | | climb as pO2 | pinned by the acceptor | pO2 rises | | drops | (flat line, slope 0) | | | slope ~ -1/6 | e' ~ -1/4 h* ~ +1/4 | slope ~ +1/6 | |------------------|--------------------------|------------------| | conducts | conducts OXIDE IONS | conducts | | ELECTRONS | (the fuel-cell window) | HOLES | Each population is a straight line on log-log; its SLOPE names the mechanism. The flat middle is why YSZ works across a wide pO2 range.
That picture is a Brouwer diagram, and its power is that each slope is a signature. Take the reducing wing of an undoped oxide, where the crystal's only charge balance is n = 2 times [V(O)]. Substitute that into K = [V(O)] times n^2 times (pO2)^(1/2) and both n and [V(O)] fall as (pO2)^(-1/6) — a straight line of slope -1/6 that shouts 'oxygen vacancies with electron compensation'. In the middle, the acceptor dopant pins [V(O)] to a constant (one vacancy per two Y3+), so the vacancy line goes flat: the oxide-ion conductivity stops caring about the gas. That plateau is exactly why an yttria-stabilized-zirconia electrolyte holds a steady oxide-ion conductivity across the huge oxygen-pressure swing between a fuel-cell's fuel side and its air side.
Clumps, and the Electrons Along for the Ride
Those clean straight lines rest on one assumption worth naming out loud: that defects are dilute and independent, drifting past each other like strangers in an empty hall. Crowd the hall and it breaks. Oppositely charged defects attract, and at real doping levels they pair up or gather into clusters — defect association. A Y(Zr)' and the V(O)** it created often bind, quietly pulling that vacancy out of traffic; this is why zirconia's oxide-ion conductivity peaks near 8 mol% yttria and then falls if you dope harder, the extra vacancies simply clumping instead of carrying current. Fe(1-x)O takes it to an extreme, its vacancies and iron interstitials condensing into ordered Koch-Cohen clusters, and heavily reduced TiO2 sheds its vacancies into crystallographic shear planes (the Magneli phases) rather than scattering them at random. The dilute model is a first approximation, honest only near it.
Look again at those doping and reduction reactions and you will see the electrons and holes were never bystanders — they are riders bolted to the ionic defects. When TiO2 lost oxygen, the electrons it freed did not float away; they settled onto titanium, turning Ti4+ into Ti3+, and hop from one such ion to the next dragging a small dimple of lattice distortion with them — a small polaron, the slow hopping conduction of many oxides. When an electron gets trapped in an empty oxygen site instead of on a cation, it can absorb visible light and colour a clear crystal — a color center, the reason additively-coloured MgO glows and irradiated crystals darken. These electronic passengers are the entire subject of guide 5; here just hold the link — every ionic defect you engineer arrives with an electronic shadow.
Step back and this guide has handed you a design language. A dopant's valence sets how many vacancies exist; the furnace's oxygen sets whether they stay ionic or turn electronic; association sets when the neat rules bend. That is the recipe behind an oxygen sensor, a fuel-cell electrolyte, a PTC thermistor, and a varistor alike. And it points straight ahead: vacancies are the empty seats atoms move through, so the populations you set here become the diffusion coefficients of the next rung, and diffusion is what welds powder into a dense part when you fire it. Author the defects well, and you are already authoring the diffusion and the sintering to come.