Gains:
- Ability to interpret phase diagram, phase ratio and transformation temperatures in a structured way with AI and query CALPHAD logic
- Ability to calculate and make sense of concepts such as leverage rule, eutectic and eutectoid transformation with AI support
- Ability to verify phase and temperature values given by AI with thermodynamic plausibility and experimental diagram
Whether to anneal a steel below or above 727°C, when an alloy will begin to solidify in casting, at what temperature a solder will completely melt—it's all written down in phase diagrams. Phase diagram is a map showing which phases (solid, liquid, different crystal structures) are in equilibrium at what temperature and composition in a material. Phase is the part of the material that is chemically and structurally uniform; for example, ferrite and cementite are two separate phases in steel. The language of metallurgy speaks with phase diagrams, and in this unit you will learn how to use artificial intelligence (systems that process text, numbers and logic) when interpreting phase diagrams, constructing leverage rule calculations, and querying CALPHAD (CALculation of PHase Diagrams: a method that calculates phase equilibrium from thermodynamic data) logic. The central principle remains the same: AI helps you read the diagram, but any temperature and composition values it gives must be verified by the empirical diagram and thermodynamic plausibility.
Basic concepts of phase diagram and role of AI
To read a phase diagram correctly, it is necessary to clarify several concepts; AI is powerful at teaching and reminding:
- Liquidus line: The temperature limit above which everything is liquid. The temperature at which the first solid begins to collapse on cooling.
- Solidus line: The boundary below which everything is solid. The temperature at which the first liquid forms upon heating.
- Eutectic transformation: The point at which a liquid simultaneously transforms into two solid phases at constant temperature (liquid → solid A + solid B). Solder alloys are the classic example of this.
- Eutectoid transformation: The point at which a solid transforms into two different solid phases at constant temperature (solid → solid A + solid B). This is the austenite → ferrite + cementite (pearlite) transformation in steel; In the iron-carbon diagram, it occurs at ~727 °C and ~0.76% carbon.
- Lever rule: A method that calculates the amount of each phase at a given temperature in a two-phase region.
When you describe a phase diagram to AI (e.g. "what phases are steel in at 800 °C with 0.4% C on the iron-carbon diagram") it gives a conceptually correct answer. However, he sometimes misremembers or makes up (hallucinates) specific temperature and composition values. So use AI as the tool that “establishes the logic of interpreting the diagram”; read the numbers from the standard diagram.
Leverage rule: step by step and verification
The lever rule gives the phase ratio in the two-phase region. Let the two phases be α and L (liquid), the total composition of the alloy be C₀, and the phase boundary compositions at that temperature be Cα and CL. According to the leverage rule:
- Mass fraction of α phase = (CL − C₀) / (CL − Cα)
- Mass fraction of phase L = (C₀ − Cα) / (CL − Cα)
When making AI do this calculation, construct the steps as follows:
- Give temperature and composition: "Cu-Ni alloy containing 40% Ni at 1450 °C."
- YOU read the phase boundary combinations: Read the Cα and CL values from the diagram yourself and give them to the AI — do not rely on what the AI "reads".
- Have AI set up the account: Apply the formula, calculate the odds and show the steps.
- Check physical plausibility: Are the odds between 0–1, the sum between 1, the temperature between liquidus/solidus?
Tip: The most common mistake in the lever rule is misreading the phase boundary combinations. Since the AI does not "see" the diagram, you must provide these values; AI only executes arithmetic and logic. If you enter the wrong Cα/CL, the result will be absolutely wrong.
CALPHAD and thermodynamic plausibility
Instead of drawing phase diagrams one by one by experiment, the CALPHAD method calculates phase equilibrium by modeling the Gibbs free energy of each phase as a function of composition and temperature and minimizing the total free energy of the system. Software such as Thermo-Calc, FactSage, Pandat do this. AI helps explain CALPHAD concepts and interpret a result; but AI itself is not a thermodynamic solver — it doesn't actually do Gibbs energy minimization, it produces the answer that looks most probable.
Correct usage: Ask the AI to interpret the CALPHAD output (e.g. a phase ratio table, a transformation temperature), flag discrepancies and query which values deviate from what is physically expected. Misuse: Telling the AI "calculate the phase diagram of this alloy" and thinking the numbers are the result of CALPHAD.
Quest
The role of AI
verification source
Explain the concept of phase diagram
Powerful — teaches
Textbook, ASM diagrams
Leverage rule arithmetic
Helper — sets up the account
Boundary combinations are read manually
Giving conversion temperature
Weak – hallucinates
Experimental/standard diagram
Interpreting the CALPHAD result
Helper — consistency check
Thermo-Calc/FactSage output
Diagram "calculation" for the new system
Inadequate – unable
Real CALPHAD software + experiment
three mini cases
Case 1 — Made-up eutectoid temperature. A student asks the AI "what is the eutectoid temperature in a steel with chromium added"; AI says "727 °C" in a determined tone. Whereas alloying elements shift the eutectoid point: chromium raises the eutectoid temperature and lowers the eutectoid carbon content. 727 °C is correct for pure iron-carbon, but varies in alloy steel. The student sees the difference when he confirms the value from the Fe-Cr-C ternary diagram or the CALPHAD calculation. Lesson: AI may incorrectly transfer the "memorized" value of the binary system to the alloy system.
Case 2 — Correctly constructed leverage calculation. An engineer wants to calculate phase ratios in a Cu-Ni alloy at 1300 °C. From the diagram, he reads the solidus composition as 32% Ni and the liquidus composition as 45% Ni; The alloy is 40% Ni. It gives these three values to the AI and applies the leverage rule. AI calculates the solid ratio as (45−40)/(45−32) = 0.38 and the liquid ratio as 0.62. The engineer checks that the sum is 1.0 and that 40% falls between the two limits. The result is reasonable and verified because the boundary combinations are read from the real diagram.
Case 3 — CALPHAD discrepancy capture. An R&D team calculates the γ′ (gamma-exponent: precipitate phase that gives high-temperature strength) ratio for a nickel superalloy in Thermo-Calc; 55% subtracts. They made the AI ask this value "is this reasonable?" AI indicates that in typical turbine superalloys the γ′ ratio is usually in the range of 40–70%, but 55% is so sensitive to database selection and temperature that it must be rerun with a different thermodynamic database. The team recalculates with two databases and compares the value with experimental image analysis. AI is correctly used here as a consistency checker.
Copiable prompt templates
PHASE DIAGRAM INTERPRETATION TEMPLATE"Role: You are a physical metallurgy assistant. System: [e.g. iron-carbon]. Composition: [%C]. Temperature: [°C]. Which phases are in equilibrium at this point? Explain what each phase is. DO NOT give the transformation temperatures; instead tell me which line (liquidus/solidus/eutectoid) I am looking at and which diagram I should read the values from. Temperature/composition values must be confirmed FROM EXPERIMENTAL DIAGRAM; fitting.”
LEVERAGE RULE CALCULATION TEMPLATE "Two-phase region. I read the following values from the diagram:- Total composition of the alloy C0 = [...]- Phase 1 boundary composition = [...]- Phase 2 boundary composition = [...]Calculate the mass ratio of each phase STEP BY STEP with the lever rule. Then do a physical check: are the ratios between 0-1, is the total 1? You do not 'read' the values; just with the numbers I have given calculate."
CALPHAD RESULT INTERPRETATION TEMPLATE"Interpret the following CALPHAD output: [phase ratios/temperatures].System and temperature: [...]. Are these values within the typical expected range?Which value might be overly sensitive to database choice or temperature?Mark which result should be confirmed with a second database or experiment. Do not claim absolute accuracy."
TRANSFORMATION MECHANISM EXPLANATION TEMPLATE"What is the [eutectic / eutectoid / peritectic] transformation, it takes place at constant temperature, which phases are formed? Explain through [alloy example]. Explain HOW the alloying elements can shift this point (in the direction of temperature and composition). Do not give specific numbers 'exactly'; give general trend + source of verification."
Weak prompt / Strong prompt
WEAK PROMPT: "Name the phases and temperatures in 0.4 carbon steel."
STRONG PROMPT: "Role: You are the physical metallurgy assistant. System: iron-carbon (balance). For 0.4% C steel, explain in order which phase regions it passes through during cooling (austenite → austenite+ferrite → pearlite+ferrite). Tell me which line I look at in each pass. State that the eutectoid temperature and composition are valid for PURE Fe-C, it will shift with the alloying element. The numbers are standard Fe-C. Remind me to confirm from the diagram and write down what to test experimentally.”
The weak prompt gives a superficial and possibly contrived warm response that overlooks the alloy effect. Powerful prompt; It requires the order of transformation logically, recalls the alloy effect, and requires confirmation of the numbers from the diagram.
Common mistakes
- Carrying the rote values of the binary system (e.g. 727 °C eutectoid) to alloy steel as is.
- "Making the AI read" the phase boundary combinations in the lever rule; AI does not see the diagram.
- Mistaking the AI for a real CALPHAD solver and accepting the resulting numbers as calculated values.
- Applying the equilibrium diagram to rapid cooling (non-equilibrium, e.g. martensite); The phase diagram shows the equilibrium state.
- Confusing liquidus/solidus with solidification interval.
- Accepting the result without checking the sum of the odds and the 0–1 limit.
In summary
Phase diagrams are the map of metallurgy; AI is powerful in teaching the logic of reading this map (phases, liquidus/solidus, eutectic/eutectoid, leverage rule) and setting up the calculation. However, it can hallucinate specific temperature and composition values, bypass the alloying effect, and is not a true CALPHAD solvent. Read the boundary combinations from the diagram, confirm the transformation values with the experimental diagram and thermodynamic software, do not apply the equilibrium diagram to non-equilibrium transformations.
Application task
Select a steel that is close to 0.8% C (near eutectoid) from the iron-carbon diagram. Ask the AI to explain with the "PHASE DIAGRAM INTERPRETATION" template what phases are present just above and just below 727 °C and why pearlite forms. Then, in a two-phase region, read the boundary combinations from the diagram yourself, have the phase ratios calculated using the "LEVERAGE RULE" template and verify that the total is 1.0. Finally, summarize in one paragraph how the alloying elements will shift the eutectoid point.
checklist
- [ ] I clarified the concepts of phase, liquidus, solidus, eutectic and eutectoid.
- [ ] I read the phase boundary combinations from the diagram myself, I did not let the AI read it.
- [ ] I checked that the leverage rule result is in the range 0–1 and sums to 1.0.
- [ ] I confirmed the transformation temperatures given by AI with the experimental diagram.
- [ ] I took into account that the alloying element shifts the balance points.
- [ ] I used AI as an interpretation and consistency tool, not a CALPHAD solver.