Maps for alloy makers
For most of the twentieth century, electronics were held together by a solder of about 63 % tin and 37 % lead. Pure lead melts at 327 °C and pure tin at 232 °C, yet this mixture melts at only 183 °C, lower than either of its ingredients, and it goes from solid to liquid at a single temperature, like a pure metal, instead of turning mushy over a range. Both properties made it ideal for soldering components without cooking them. When the European Union restricted lead in electronics from 2006, the industry needed another alloy with the same trick, and found it in tin with a few per cent of silver and copper, which melts near 217 °C.
Where does a "lowest melting" composition come from, and how would anyone find the next one? The answer is a phase diagram: a map that shows, for every composition and temperature, which phases are present at equilibrium, what each is made of and how much of each there is. Phase diagrams tell a steelmaker which temperature turns a steel into austenite, a foundry why one aluminium alloy pours well, a road engineer why salt stops melting ice in very cold weather, and a chip-maker which metals will react with silicon. They are the most used reference in materials engineering.
This lesson reads phase diagrams and explains where they come from. It begins with definitions and Gibbs's phase rule, works through the three basic kinds of binary diagram, reads the iron–carbon diagram that governs all steels, and ends by computing the free-energy curves that the diagrams are drawn from.
Phases, components and equilibrium
Definition— component
The components of a system are the chemically independent constituents needed to describe its composition: Cu and Ni for copper–nickel alloys, Fe and C for steels, H₂O and NaCl for salty water. A system of two components is binary.
Definition— phase
A phase is a region of material that is uniform in structure and composition, separated from other phases by boundaries. Ice, liquid water and water vapour are three phases of one component. So are ferrite (body-centred cubic iron) and austenite (face-centred cubic iron): two solids with different crystal structures are different phases. A solid solution with dissolved atoms on lattice sites or in the gaps between them is still one phase.
Definition— equilibrium and the phase diagram
At constant temperature and pressure, a system is in equilibrium when its Gibbs energy is as low as it can be. An equilibrium phase diagram shows, against temperature and composition (at a fixed pressure, usually 1 atm), which phases have the lowest total . It does not say how long the system takes to get there; that is the subject of phase transformations.
Lumenary Chemistry's gases, liquids and solids reads one-component diagrams such as water's, plotted against pressure and temperature. Materials scientists mostly work at atmospheric pressure, where the pressure of a solid hardly matters, and want to know what happens when they mix. So their diagrams put composition across and temperature up. How many variables can be changed independently while a given set of phases stays in equilibrium? Josiah Willard Gibbs answered in the 1870s.
Theorem— The Gibbs phase rule
For a system of components with phases in equilibrium, the number of intensive variables (temperature, pressure, compositions) that can be varied independently without changing the number of phases is
At fixed pressure, as on the diagrams in this lesson, .
Proof
Each phase's composition is fixed by mole fractions (they sum to 1), so with temperature and pressure there are intensive variables. Equilibrium requires each component to have the same chemical potential in every phase, as the free-energy section below explains: equations for each of the components, in all. The number of free variables is the difference, . Fixing the pressure removes one more.
∎
For a binary alloy at fixed pressure, . A single phase has : temperature and composition can both change, so single phases occupy areas of the diagram. Two phases have : choose the temperature and the compositions of both phases are fixed, which is why two-phase regions are crossed by horizontal tie lines. Three phases have : they can coexist only at one temperature, with fixed compositions, so three-phase equilibria appear as horizontal lines called invariant reactions. A binary diagram can never show four phases in equilibrium.
Complete solubility and the lever rule
The simplest binary diagram belongs to two metals that dissolve in each other completely, in the solid as well as the liquid. Copper and nickel are the classic pair: both are face-centred cubic, their atoms differ in size by about 2 %, and they have similar electronegativities and valences. (These are the conditions William Hume-Rothery identified for wide solid solubility, discussed further in metals and alloys.) Their diagram has just two curves: the liquidus, above which everything is liquid, and the solidus, below which everything is solid. Between them, solid and liquid coexist, and because the shape is a lens it is often called a lens diagram. Such a system is isomorphous.
The shape of the lens can be computed. If both the liquid and the solid are ideal solutions, a component's chemical potential in a phase is its pure value plus , and equating chemical potentials between solid and liquid for each element gives , with . Two such equations, one per element, fix both compositions at each temperature. The block computes the lens from rounded handbook values of the melting points and enthalpies of fusion of copper and nickel, then applies the lever rule, stated after the block, to an alloy with nickel mole fraction 0.44 at 1250 °C.
import numpy as np
R = 8.314 # J/(mol K)
# Melting points (K) and enthalpies of fusion (J/mol): rounded handbook values.
Tm = {"Cu": 1357.8, "Ni": 1728.0}
Hf = {"Cu": 13.26e3, "Ni": 17.48e3}
def lens(T):
"""Mole fractions of Ni in the solid and the liquid in equilibrium at T (K),
if both phases are ideal solutions: x_liquid / x_solid = exp(-dG_fusion / RT) for each element."""
kCu = np.exp(-Hf["Cu"] * (1 - T / Tm["Cu"]) / (R * T))
kNi = np.exp(-Hf["Ni"] * (1 - T / Tm["Ni"]) / (R * T))
x_solid = (1 - kCu) / (kNi - kCu)
return x_solid, kNi * x_solid
print(" T (C) solidus x_Ni liquidus x_Ni")
for T_c in range(1100, 1451, 50):
xs, xl = lens(T_c + 273.15)
print(f"{T_c:5d} {xs:12.3f} {xl:13.3f}")
# The lever rule: an alloy with x_Ni = 0.44 held at 1250 C.
x0, T = 0.44, 1250 + 273.15
xs, xl = lens(T)
f_solid = (x0 - xl) / (xs - xl)
print(f"\nat 1250 C the tie line runs from liquid x_Ni = {xl:.3f} to solid x_Ni = {xs:.3f}")
print(f"an alloy with x_Ni = {x0} is {100 * f_solid:.1f} % solid and {100 * (1 - f_solid):.1f} % liquid (by moles)")
# Where does this alloy start and finish freezing? Bisection on the liquidus and solidus.
def crossing(which):
lo, hi = Tm["Cu"] + 0.01, Tm["Ni"] - 0.01 # x_Ni along each curve rises with T
for _ in range(60):
mid = (lo + hi) / 2
x = lens(mid)[which]
lo, hi = (mid, hi) if x < x0 else (lo, mid)
return mid - 273.15
print(f"it starts to freeze at {crossing(1):.0f} C (liquidus) and is solid below {crossing(0):.0f} C (solidus)")At every temperature the solid is richer in nickel, the higher-melting metal, than the liquid it is in equilibrium with: at 1250 °C the tie line runs from liquid with to solid with . The ratio of the two, for nickel here, is the partition (or distribution) coefficient. It decides how a solute is redistributed when an alloy freezes, from the uneven composition of castings to zone refining, which purifies silicon and germanium for electronics by sweeping a molten zone along a bar so that impurities with are carried to one end. The alloy with starts to freeze at 1264 °C, is 52.5 % solid at 1250 °C, and is completely solid below 1237 °C.
Theorem— The lever rule
On a tie line from phase α of composition to phase β of composition , an alloy of overall composition consists of the fractions
The fractions are by moles when the compositions are mole fractions, and by mass when they are mass fractions.
Proof
Conservation of matter: the total amount of the alloy is , and the amount of component B is the same whether counted in the alloy or in the two phases, . Substituting and solving gives the result. Geometrically, : the alloy point is the balance point of a lever with masses and at its ends, which is where the name comes from.
∎
The rule is easy to get backwards. The fraction of β is the length of the arm on the α side divided by the whole tie line: the closer the alloy sits to one end, the more of that phase it contains. For the alloy above, .
Eutectics and other invariant reactions
Most pairs of metals do not dissolve in each other completely in the solid. Lead dissolves at most about 18 wt% tin, and tin barely any lead, because their crystal structures differ. Their diagram has two solid phases, α (lead-rich, face-centred cubic) and β (tin-rich, tetragonal), each a single phase over a narrow range of composition, with a two-phase field α + β between them. The two liquidus lines fall from the melting points of the pure metals and meet at the eutectic point, from the Greek for "easily melted".
At the eutectic point three phases coexist, so and the eutectic reaction
happens at exactly 183 °C. A liquid of the eutectic composition freezes completely at that one temperature into a fine mixture of the two solids, often as alternating plates (lamellae) a few micrometres or less apart, because atoms only have to diffuse sideways from one plate to the next. The eutectic is not a compound: it is a microconstituent, a recognisable arrangement of two phases.
Example— A 40 wt% Sn solder cooling through the eutectic
An alloy with 40 wt% tin starts to freeze when it crosses the α liquidus, at about 240 °C, and α crystals grow as it cools. Just above 183 °C, the tie line runs from α (18.3 wt% Sn) to liquid (61.9 wt% Sn), so
At 183 °C that remaining liquid, which now has exactly the eutectic composition, freezes as eutectic. Just below 183 °C the phases are α and β, with and . So the solid contains 50.2 % primary α, which formed first as rounded or tree-like grains, and 49.8 % eutectic, which itself is 22.5 % (of the whole) α and 27.3 % β. A metallographer polishing and etching the solder would see large α grains in a fine striped matrix, and could check these numbers by measuring the area fractions. The lever rule gives the amounts of phases or of microconstituents, depending on which tie line is used.
Three-phase reactions come in a few kinds, all horizontal lines on a binary diagram. In each, one phase on cooling becomes two, or two become one.
| Reaction | On cooling | Example |
|---|---|---|
| eutectic | L → α + β | Pb–Sn at 183 °C; Al–Si casting alloys at 577 °C |
| eutectoid | γ → α + β (all solid) | austenite → ferrite + cementite in steel at 727 °C |
| peritectic | L + α → β | in Fe–C near 1495 °C; in many copper alloys |
Diagrams can also contain intermediate compounds such as Fe₃C, Mg₂Si or the intermetallics in solder joints, which appear as vertical lines or narrow single-phase fields and split a complicated diagram into simpler pieces.
The iron–carbon diagram
Steel is iron with up to about 2 wt% carbon, and its heat treatments are planned on the iron–carbon diagram. Pure iron is body-centred cubic ferrite (α) up to 912 °C, face-centred cubic austenite (γ) from 912 to 1394 °C, and body-centred cubic again (δ) up to its melting point at 1538 °C. Carbon sits in the gaps between iron atoms (see crystal structures), and the gaps in face-centred cubic iron are larger, so austenite dissolves up to 2.14 wt% carbon while ferrite dissolves at most 0.022 wt%. Beyond its solubility, carbon forms cementite, Fe₃C, a hard, brittle carbide with 6.70 wt% carbon.
The most important feature is the eutectoid at 727 °C and 0.76 wt% carbon, where austenite becomes ferrite plus cementite: . Because the reaction happens in the solid, it produces a very fine lamellar mixture called pearlite (it looks like mother-of-pearl under the microscope). A steel with exactly 0.76 wt% carbon, slowly cooled, is all pearlite. A hypoeutectoid steel with less carbon first grows ferrite at the austenite grain boundaries as it cools from the A₃ line to 727 °C, and the remaining austenite, enriched to 0.76 wt%, then becomes pearlite. A hypereutectoid steel grows cementite first, below the line. Steels above 2.14 wt% carbon would meet a eutectic at 1147 °C and 4.30 wt% carbon; those are the cast irons, which melt at lower temperatures and pour well.
The block applies the lever rule just below 727 °C across the steel range, both to microconstituents (proeutectoid phase and pearlite) and to phases (total ferrite and cementite).
# The steel corner of the iron-carbon (Fe-Fe3C) diagram, just below the eutectoid at 727 C.
# Compositions in wt% carbon (assessments differ slightly: the eutectoid is quoted as 0.76-0.77).
C_ferrite, C_eutectoid, C_cementite = 0.022, 0.76, 6.70
def lever(c, left, right):
"""Fraction of the right-hand phase for overall composition c on a tie line left-right."""
return (c - left) / (right - left)
print(" wt% C proeutectoid pearlite total ferrite total cementite")
for c in (0.05, 0.20, 0.40, 0.60, 0.76, 1.00, 1.40):
if c <= C_eutectoid: # hypoeutectoid: ferrite forms first, between A3 and 727 C
pearlite = lever(c, C_ferrite, C_eutectoid)
first = f"ferrite {1 - pearlite:5.3f}"
else: # hypereutectoid: cementite forms first, on the grain boundaries
pearlite = 1 - lever(c, C_eutectoid, C_cementite)
first = f"cementite {1 - pearlite:3.3f}"
cementite = lever(c, C_ferrite, C_cementite)
print(f"{c:6.2f} {first:>16} {pearlite:8.3f} {1 - cementite:13.3f} {cementite:15.3f}")
inside = lever(C_eutectoid, C_ferrite, C_cementite)
print(f"\npearlite itself is {100 * inside:.1f} % cementite and {100 * (1 - inside):.1f} % ferrite by mass")
# The eutectoid composition in atom per cent, for comparison with cementite, Fe3C (25 at% C).
M_C, M_Fe = 12.011, 55.845
for name, w in (("eutectoid", C_eutectoid), ("cementite", C_cementite)):
nC, nFe = w / M_C, (100 - w) / M_Fe
print(f"{name:9}: {w:.2f} wt% C = {100 * nC / (nC + nFe):.1f} at% C")A 0.40 wt% carbon steel, typical of shafts and bolts, is about half proeutectoid ferrite and half pearlite (0.488 and 0.512 by mass), and only 5.7 % of it is cementite. Pearlite itself is 11.1 % cementite, so its lamellae of ferrite are about eight times thicker than those of cementite. As carbon rises the fraction of hard cementite rises in proportion, which is why strength and hardness increase and ductility falls; past the eutectoid, the proeutectoid cementite forms a brittle network along the old austenite grain boundaries, which is why plain high-carbon steels need careful heat treatment. The eutectoid composition is only 3.4 atom per cent carbon: a little carbon goes a long way.
Free-energy curves behind the diagrams
Where do the lines come from? At each temperature, every possible phase has a molar Gibbs energy that depends on composition, . Lumenary Chemistry's entropy and free energy explains why equilibrium is where is lowest, and its chemical equilibrium applies the idea to reacting mixtures. For a mixture, though, "lowest" has a twist: an alloy need not be one phase. It can split into two phases of different compositions, and its Gibbs energy is then the lever-rule average of theirs, which lies on the straight line joining the two points on the curves.
Theorem— The common-tangent construction
At fixed temperature and pressure, two phases with compositions and are in equilibrium when one straight line is tangent to the curve at and to at . For any overall composition between them, the two-phase mixture has a lower Gibbs energy than any single phase. The phase boundaries of the diagram at that temperature are the tangent points.
Proof
For a two-phase mixture of overall composition , with the lever-rule fractions, which is the value at of the straight line through and . Minimising over the choice of and means lowering that chord as far as possible; it reaches its lowest position when it touches both curves without crossing either, that is, when it is tangent to both. The tangent's intercepts at and are the chemical potentials of the two components, so the condition is the same as equal chemical potentials of each component in both phases.
∎
The simplest model that produces a two-phase region is the regular solution. Mixing atoms A and B adds an ideal entropy of mixing, which always favours mixing, and an enthalpy of mixing , where the interaction parameter is positive when A–B bonds are weaker than the average of A–A and B–B bonds:
At high temperature the entropy term wins and is a single bowl: one solution at every composition. Below the critical temperature the curve develops two minima with a hump between them, and compositions under the hump split into two phases: a miscibility gap. Inside the gap there is a smaller region where , bounded by the spinodal, where the solution is unstable even to tiny fluctuations and separates without needing to nucleate, which is spinodal decomposition (see phase transformations). The block computes the gap for kJ mol⁻¹.
import numpy as np
R = 8.314 # J/(mol K)
W = 20e3 # interaction parameter Omega, J/mol (positive: A-B bonds are unfavourable)
def G_mix(x, T):
"""Gibbs energy of mixing of a regular solution, J/mol."""
return W * x * (1 - x) + R * T * (x * np.log(x) + (1 - x) * np.log(1 - x))
def binodal(T):
"""Left edge of the miscibility gap: for this symmetric model the common tangent is
horizontal, so solve dG/dx = W(1 - 2x) + RT ln(x/(1-x)) = 0 for 0 < x < 1/2 by bisection."""
lo, hi = 1e-12, 0.5 - 1e-9
for _ in range(200):
mid = (lo + hi) / 2
slope = W * (1 - 2 * mid) + R * T * np.log(mid / (1 - mid))
lo, hi = (lo, mid) if slope > 0 else (mid, hi)
return mid
def spinodal(T):
"""Where d2G/dx2 = RT/(x(1-x)) - 2W = 0, i.e. x(1-x) = RT/(2W)."""
return 0.5 - np.sqrt(0.25 - R * T / (2 * W))
Tc = W / (2 * R)
print(f"critical temperature Tc = W/2R = {Tc:.1f} K\n")
print(" T (K) gap from x = to x = spinodal from to")
for T in (600, 800, 1000, 1100, 1150, 1200):
b, s = binodal(T), spinodal(T)
print(f"{T:5d} {b:12.4f} {1 - b:6.4f} {s:13.4f} {1 - s:6.4f}")
# Below Tc a solution inside the gap lowers its Gibbs energy by splitting into two phases.
# The common tangent touches the curve at the two binodal compositions; the alloy's
# Gibbs energy as a two-phase mixture lies on that line.
T = 900
b = binodal(T)
tangent = G_mix(b, T) # horizontal here, by symmetry
print(f"\nat {T} K the common tangent touches G_mix at x = {b:.3f} and {1 - b:.3f}, at {tangent:.0f} J/mol")
print(" x one solution (J/mol) two phases (J/mol) gain from splitting")
for x in (0.15, 0.20, 0.30, 0.40, 0.50):
g = G_mix(x, T)
print(f"{x:5.2f} {g:20.0f} {tangent:18.0f} {g - tangent:19.0f}")The critical temperature is 1202.8 K. At 600 K the gap spans almost the whole range, from to ; it narrows as the temperature rises and closes at , giving the dome in the figure. At 900 K a 50:50 solution would have J mol⁻¹, but by splitting into two phases with and it reaches J mol⁻¹, a gain of 448 J mol⁻¹. Real systems with gaps of this kind include gold–nickel and the aluminium–zinc alloys, and the same physics, with written in terms of polymer–polymer contacts, explains why most pairs of polymers do not mix.
Reading real diagrams
Real phase diagrams are measured and then assessed: specialists collect every experiment on a system, from thermal analysis, X-ray diffraction of quenched samples and microscopy, and fit a consistent thermodynamic description. The standard collections are ASM International's Alloy Phase Diagram Database and its handbooks for metals, and the NIST/ACerS Phase Equilibria Diagrams (NIST Standard Reference Database 31) for ceramics; both are in the references. A few habits make reading them safer:
- Check the units and the axis. Diagrams are drawn in weight per cent or atom per cent, often with the other scale along the top. The difference is large for elements of very different mass: 0.76 wt% carbon in iron is 3.4 at%.
- Check whether lines are measured or estimated. Assessments draw uncertain boundaries dashed. Low-temperature solvus lines are often uncertain because equilibrium there takes years to reach.
- Check stable or metastable. The iron–cementite diagram is metastable; many diagrams in handbooks show both.
- For three or more components, use sections. Real alloys have many elements. Ternary diagrams are shown as triangular isothermal sections at one temperature, or as isopleths (vertical sections at one fixed ratio), which do not obey the lever rule because tie lines leave the section.
Since the 1970s, following Larry Kaufman's work, diagrams have increasingly been calculated by the CALPHAD method. Each phase gets a model of with parameters fitted to experiments and to first-principles calculations, stored in databases, and the equilibrium is found by minimising the total , which is the common-tangent construction done numerically in many dimensions. Commercial packages such as Thermo-Calc (Sweden), FactSage (Canada and Germany) and Pandat (United States), and open-source codes such as OpenCalphad and pycalphad, do this for alloys of a dozen elements. CALPHAD made it practical to explore high-entropy alloys, introduced in 2004 by groups in the United Kingdom and Taiwan: alloys of five or more elements in near-equal amounts, which live in the middle of multicomponent diagrams that nobody had mapped. As of 2026, first-principles and machine-learned calculations are feeding more and more of the parameters in these databases, especially for systems where experiments are scarce (see AI-assisted materials discovery).
Common misconceptions
Misconception
"The lever rule gives the compositions of the phases." The ends of the tie line give the compositions; the lever rule gives the amounts. And the amount of a phase is the length of the opposite arm divided by the whole tie line.
Misconception
"A phase diagram shows the microstructure you will get." It shows equilibrium. A steel quenched from austenite forms martensite, which is not on the diagram at all, and a casting cooled in minutes is cored. The diagram says where the system is heading; kinetics, the subject of the next lesson, says how far it gets.
Misconception
"A eutectic is a compound with its own formula." A eutectic is a fine mixture of two phases that forms at one composition and temperature. Its composition is set by the shapes of the free-energy curves, not by a ratio of whole numbers, and it changes, slightly, with pressure.
Misconception
"A phase means solid, liquid or gas." Those are states of matter. A solid can contain several phases, such as ferrite and cementite in pearlite, and some systems have two liquid phases, like oil and water.
Exercises
For the iron–carbon exercises, use 0.022, 0.76 and 6.70 wt% carbon for ferrite, the eutectoid and cementite. Take J mol⁻¹ K⁻¹.
1. Phase diagrams
Warm-upHumanA binary alloy at atmospheric pressure sits exactly at its eutectic point, with liquid, α and β in equilibrium. How many degrees of freedom does the Gibbs phase rule allow?
2. Phase diagrams
CoreHumanIn an isomorphous alloy system at 1300 °C, the tie line runs from liquid with nickel mole fraction 0.54 to solid with nickel mole fraction 0.61. An alloy with overall nickel mole fraction 0.59 is held at 1300 °C until it reaches equilibrium. What percentage of it (by moles) is solid? Give the nearest whole number.
3. Phase diagrams
CoreHumanA steel with 0.30 wt% carbon is cooled slowly from the austenite field to just below 727 °C. What percentage of it, by mass, is pearlite? Give the nearest whole number. (Ferrite 0.022, eutectoid 0.76 wt% C.)
4. Phase diagrams
CoreHumanA steel with 0.40 wt% carbon is held at 760 °C until it reaches equilibrium. Using the iron–carbon diagram, which phases are present?
5. Phase diagrams
Warm-upHumanA binary solid solution follows the regular-solution model with interaction parameter kJ mol⁻¹. Above what temperature, in kelvin to the nearest whole number, do the two components mix in all proportions?
6. Phase diagrams
CoreHumanWhich statements about an alloy of exactly the eutectic composition (for example 61.9 wt% tin in lead) are true? Select all that apply.
7. Phase diagrams
CoreComputationalWrite steel_constituents(c) for a plain carbon steel with c wt% carbon, slowly cooled to just below 727 °C. It returns a tuple (name, f_proeutectoid, f_pearlite): the proeutectoid phase ("ferrite" for , "cementite" above), its mass fraction, and the mass fraction of pearlite. Use 0.022, 0.76 and 6.70 wt% carbon for ferrite, the eutectoid and cementite. Raise ValueError if c is outside the steel range 0.022 to 2.14 wt%.
8. Phase diagrams
StretchComputationalWrite miscibility_gap(W, T, R=8.314) for a regular solution with interaction parameter W (J/mol) at temperature T (K). Return the gap's edges (x1, x2) with x1 < 0.5 < x2, or None if T is at or above the critical temperature . Because the model is symmetric, the common tangent is horizontal and the edges solve ; find the root in by bisection, and use symmetry for the other edge.
Further reading and resources
- DoITPoMS: Phase diagrams and solidification · University of CambridgeA free teaching package with interactive binary diagrams and the lever rule.
- DoITPoMS: Ternary phase diagrams · University of Cambridge
- ASM InternationalPublishes the ASM Alloy Phase Diagram Database and the ASM Handbook volume on alloy phase diagrams, the standard sources of assessed diagrams.
- NIST Standard Reference Database 31: NIST/ACerS Phase Equilibria Diagrams · National Institute of Standards and Technology and The American Ceramic SocietyPhase diagrams for ceramic and other non-metallic systems.
- The C-Fe (carbon-iron) system · H. Okamoto · 1992Journal of Phase Equilibria. An assessment of the iron–carbon diagram.
- Computational Thermodynamics: The Calphad Method · Hans Leo Lukas, Suzana G. Fries and Bo Sundman · 2007Cambridge University Press. How phase diagrams are calculated from thermodynamic models.
- pycalphad: CALPHAD-based Computational Thermodynamics in Python · Richard Otis and Zi-Kui Liu · 2017Journal of Open Research Software. An open-source library that computes phase diagrams from thermodynamic databases; documentation at pycalphad.org.
- Microstructural development in equiatomic multicomponent alloys · B. Cantor, I. T. H. Chang, P. Knight and A. J. B. Vincent · 2004Materials Science and Engineering A. One of the two 2004 papers that started the study of high-entropy alloys.
- Nanostructured high-entropy alloys with multiple principal elements · J.-W. Yeh and co-workers · 2004Advanced Engineering Materials. The other founding paper, which named high-entropy alloys.
- Materials Project documentationExplains the computed phase diagrams (convex hulls) built from density functional theory.