Deoxidation equilibrium predicts the dissolved oxygen activity that can coexist with a specified deoxidizer activity and oxide phase at a defined temperature. It does not directly predict total oxygen, inclusion removal, or the alloy addition required in a ladle. A credible calculation must identify the oxide product, use consistent thermodynamic standard states, and account for interactions among dissolved elements.

The practical consequence is significant: two calculations using the same aluminum concentration can predict different oxygen levels because they assume different alumina activities, steel compositions, or temperatures. Comparing the oxygen affinity of aluminum, silicon, and manganese is therefore only a starting point. The decisive question is which metal–oxide equilibrium is physically compatible with the steel and its surrounding slag.

Define the equilibrium before selecting thermodynamic data

For a deoxidizer M forming an oxide M_xO_y, write the reaction explicitly:

x[M] + y[O] ⇌ M_xO_y

Square brackets denote species dissolved in liquid steel. The oxide may be a pure solid, a component of a liquid inclusion, or a component of the refining slag. These alternatives require different activity assumptions.

For the formation reaction as written:

ΔG = ΔG° + RT ln Q

Q = a_oxide / (a_M^xa_O^y)

At equilibrium, ΔG = 0 and Q = K, giving:

ln K = −ΔG°/(RT)

a_O = [a_oxide/(Ka_M^x)]^1/y

Here, T is absolute temperature, R is the gas constant, and all activities are dimensionless relative to their specified standard states. ΔG° and R must use compatible energy units.

Reaction direction is a frequent source of error. A reference may tabulate oxide dissociation rather than oxide formation. Reversing the reaction reverses the sign of ΔG° and replaces K with its reciprocal. Changing stoichiometric coefficients also changes the numerical equilibrium constant.

An Ellingham diagram helps compare oxide stability, but its standard states may differ from those used for solutes in liquid iron. Values based on pure elements and gaseous oxygen cannot be inserted directly into a dissolved-oxygen calculation without the appropriate thermodynamic transformations.

Activities, not concentrations, control the result

For dilute solutes in liquid iron, a common convention uses a hypothetical 1 mass-percent Henry-law standard state:

a_i = f_i[%i]

In this expression, [%i] is the numerical mass-percent concentration and f_i is its activity coefficient. A concentration of 0.03 mass% is entered as 0.03, not 0.0003. Mixing mass fractions with coefficients defined for mass-percent concentrations can invalidate the result by orders of magnitude.

The first-order Wagner interaction formalism expresses composition effects as:

log₁₀f_i = Σ e_i^j[%j]

The interaction parameter e_i^j describes how dissolved element j affects the activity coefficient of i. Both oxygen and the deoxidizer require attention: adjusting oxygen activity while assuming ideal aluminum behavior is not automatically consistent.

This approximation is useful within the dilute-composition and temperature ranges supported by its parameters. Higher-order terms or an assessed solution model may be necessary when interactions are strong or alloy concentrations move beyond that range. Extrapolating a dilute iron model into high-chromium or other highly alloyed melts requires explicit validation.

A defensible metallurgical thermodynamics guide must therefore distinguish three separate inputs: the equilibrium constant, the metallic solution model, and the oxide solution model. Accurate data in one category cannot compensate for an unsuitable assumption in another.

Aluminum deoxidation: powerful equilibrium, sensitive assumptions

For equilibrium with alumina:

2[Al] + 3[O] ⇌ Al₂O₃

K_Al = a_Al₂O₃/(a_Al²a_O³)

Consequently:

log₁₀a_O = ⅓(log₁₀a_Al₂O₃ − log₁₀K_Al − 2log₁₀a_Al)

Setting alumina activity to unity is appropriate for equilibrium with pure solid alumina in its chosen standard state. It is not a universal assumption for every aluminum-bearing inclusion. Alumina dissolved in a calcium-aluminate liquid or another multicomponent oxide phase requires an activity obtained from a suitable oxide model.

The stoichiometry provides useful sensitivity checks without requiring a particular thermodynamic dataset:

• Doubling aluminum activity lowers equilibrium oxygen activity to 2^−2/3, approximately 0.63 of its initial value, if temperature and alumina activity remain fixed.
• Reducing alumina activity from 1 to 0.1 lowers equilibrium oxygen activity to approximately 0.46 of its initial value, with other conditions unchanged.

These are activity relationships, not guaranteed concentration responses. If aluminum addition changes f_O or f_Al, the dissolved concentrations must be recalculated. Likewise, an assumed reduction in alumina activity is meaningful only if the proposed oxide composition and phase assemblage can actually exist.

The relevant aluminum input is dissolved aluminum, not aluminum bound in inclusions. Analytical categories such as acid-soluble aluminum are method-dependent and should not be treated as exact thermodynamic species definitions without checking the measurement procedure.

Silicon and manganese require a coupled oxide calculation

The simplified reactions are:

[Si] + 2[O] ⇌ SiO₂

[Mn] + [O] ⇌ MnO

The corresponding oxygen activities are:

a_O = [a_SiO₂/(K_Sia_Si)]^1/2

a_O = a_MnO/(K_Mna_Mn)

Evaluating both reactions against separate pure oxides can misrepresent combined silicon–manganese deoxidation. A manganese-silicate inclusion has component activities determined by its composition and non-ideal mixing behavior. Oxide mole fraction alone is not generally an adequate substitute for activity.

Combining the reactions gives an additional consistency relation:

[Si] + 2MnO ⇌ SiO₂ + 2[Mn]

With consistent standard states, its equilibrium constant is K_Si/K_Mn². This relation constrains the partitioning of manganese and silicon between the steel and oxide phase. A solution that independently satisfies each reaction using incompatible oxide activities is not a valid coupled equilibrium.

The engineering benefit of a mixed oxide is therefore not explained solely by ranking deoxidizers. Mixing changes component activities and may change whether the inclusion is solid or liquid. Its melting behavior and subsequent removal still require separate phase and transport assessments.

Temperature and phase stability can change the interpretation

Temperature affects the equilibrium constant, metallic activity coefficients, oxide activities, and stable phase assemblage. For a reaction with an approximately constant standard enthalpy over the interval considered:

d(ln K)/dT = ΔH°/(RT²)

For exothermic oxide formation, increasing temperature decreases the formation constant. At fixed deoxidizer and oxide activities, the equilibrium oxygen activity consequently rises. The actual concentration response also includes the temperature dependence of activity coefficients.

A ladle-temperature result should not be transferred unchanged to casting or solidification. Cooling can change oxide saturation, while solute segregation during solidification changes local concentrations. Secondary inclusion formation may therefore occur even after an earlier liquid-steel sample appeared consistent with the refining target.

Phase stability must also be checked rather than assumed. For the formation reaction, Q < K indicates a thermodynamic driving force toward oxide formation. However, precipitation may be delayed by a nucleation barrier. Conversely, a calculation that imposes equilibrium with an oxide absent from the system may describe a saturation boundary rather than the actual bath condition.

Slag–metal reactions determine whether the calculated limit is reachable

A low oxygen activity calculated for aluminum–alumina equilibrium does not ensure that the ladle will maintain it. Oxidizing slag components can consume dissolved aluminum and supply oxygen through interfacial reactions. One relevant reaction is:

2[Al] + 3(FeO) ⇌ Al₂O₃ + 3Fe

Parentheses denote slag constituents. FeO activity, not simply reported FeO mass percentage, enters the equilibrium expression. MnO and other reducible constituents can also participate, depending on the system.

There are two distinct calculation problems. A saturation calculation specifies dissolved composition and oxide activity, then calculates equilibrium oxygen. A closed-system equilibrium calculation instead specifies total element inventories and solves for metal composition, slag composition, inclusion quantities, and stable phases simultaneously.

For the latter, oxygen and deoxidizer balances are indispensable. Holding aluminum concentration fixed while permitting an unlimited quantity of oxidizing slag to react implicitly assumes an external aluminum reservoir. That boundary condition is unsuitable for estimating alloy consumption in a finite ladle.

Refractory reactions and atmospheric reoxidation can impose additional boundaries. Bulk steel, interfacial regions, and inclusions need not share one instantaneous equilibrium state. A single global equilibrium result is best treated as a limiting condition unless mixing and reaction times justify a stronger interpretation.

Translate equilibrium oxygen into a measurable quantity

After calculating oxygen activity, dissolved oxygen concentration follows from:

[%O] = a_O/f_O

Dissolved oxygen in mass ppm = 10,000 × [%O]

For example, an activity of 0.0005 would correspond to 5 mass ppm only if f_O = 1 under the stated mass-percent standard-state convention. This is a conversion example, not a predicted oxygen level for a particular steel.

Total oxygen is not interchangeable with dissolved oxygen. Total oxygen includes oxygen contained in sampled oxide inclusions. A bath close to its dissolved-oxygen equilibrium can still show substantial total oxygen if inclusions have not separated. Conversely, sampling location, entrained slag, and sample contamination can complicate the comparison.

Electrochemical oxygen probes respond to oxygen potential and require the applicable calibration and temperature interpretation. Inert-gas fusion analysis measures oxygen in the analyzed sample, including oxide-bound oxygen. Comparing either measurement with a model requires matching the reported quantity, sampling time, and temperature.

What makes a prediction technically defensible?

A reproducible calculation should identify the reaction equations, standard states, thermodynamic data source, temperature range, interaction parameters, oxide model, and permitted phases. An assessed thermochemical database or a documented steelmaking reference is preferable to combining isolated constants whose conventions are unclear.

The numerical procedure should satisfy mass balances and all active equilibrium relations together. Because activity coefficients depend on composition, iteration may be required. A Gibbs-energy minimization calculation can handle multiple phases, but only within the chemical system and phases included in its database; software cannot correct an omitted relevant phase.

Sensitivity analysis should focus on uncertain physical inputs rather than merely solver precision. Changing the oxide activity assumption, plausible bath temperature, or metallic interaction model can be more consequential than reporting another decimal place in oxygen concentration.

Finally, equilibrium should be separated from removal kinetics. Mixing, alloy dissolution, nucleation, inclusion growth, collision, flotation, and slag capture determine how rapidly the bath approaches a thermodynamic limit and whether oxide products leave the steel. Low equilibrium dissolved oxygen establishes a deoxidation potential; low total oxygen additionally requires effective inclusion separation and control of reoxidation.

The strongest assessment therefore reports a conditional result: the predicted dissolved oxygen under a specified temperature, composition, oxide assemblage, and boundary condition, accompanied by the assumptions most likely to change it. That is a more useful basis for refining decisions than an isolated “minimum oxygen” value detached from the steelmaking system.