Skip to content
LumenaryAcademy
Mechanical 80 min

Elasticity, plasticity and strengthening

Why stiffness comes from bonds and strength from dislocations, and the four ways metallurgists make a metal stronger.

Code tested · 6AI-draftedAwaiting review by a materials scientist

Your guides: Chloe · The Materials Scientist, Margaret · The Tester, Gideon · The Failure Analyst

MotivationH

Stiff, strong and hard are three different things

Take three pieces of steel: a soft iron nail, the spring from a clothes peg and a length of piano wire. Hang the same small weight from each and measure how much each stretches. To within a few per cent they stretch the same amount: every steel has a Young's modulus of about 200 GPa. Now keep adding weight. The nail's steel starts to stretch permanently at perhaps 250 MPa; the piano wire carries ten times that before anything permanent happens.

So stiffness, which sets how much a part deflects, is almost impossible to change by processing, while strength, which sets the load it bears without permanent damage, can be changed tenfold by alloying, rolling and heat treatment. Stiffness comes from the bonds between atoms, which processing hardly touches. Strength comes from dislocations, the line defects of Defects and dislocations, and from everything that gets in their way, which processing changes completely.

Lumenary ME's stress and strain treats these properties as an engineer uses them, with Hooke's law and the tensile test. Here we go underneath, from the bond to the dislocation to the alloy designer's toolkit, and finish with hardness, the quickest test of all.

IntuitionH

Springs between atoms

Two atoms bonded together sit at the bottom of an energy well. Push them closer and they repel strongly; pull them apart and they attract, more and more weakly as the gap grows. The figure shows the energy U(r)U(r) of a typical bond and, below it, the force dU/drdU/dr needed to hold the atoms at separation rr.

energy Urforce to hold the atoms apart, dU/drrr₀: bottom of the welllargest force: ideal strengthgentle side: bonds stretchedsteep side: atoms squeezedslope at r₀ = bond stiffnessattraction fades with distance
A bond's energy (top) and the force needed to hold the two atoms apart (bottom), drawn for a Lennard-Jones bond. At the equilibrium spacing r₀ the force is zero; the slope of the force curve there is the bond's stiffness, which sets the elastic moduli. The force cannot exceed a maximum, reached at a stretch of about 11%: that is the ideal strength of the bond.

Near the bottom of the well the curve is almost a parabola, so for small stretches the restoring force is proportional to the stretch: each bond is a tiny spring with stiffness S0=U′′(r0)S_0 = U''(r_0). A solid is a three-dimensional mattress of such springs. A stress σ\sigma is a force per bond divided by the area per bond, about r02r_0^2, and a strain is a stretch divided by r0r_0, so

E≈S0/r02⋅ΔrΔr/r0=S0r0.E \approx \frac{S_0/r_0^2 \cdot \Delta r}{\Delta r / r_0} = \frac{S_0}{r_0}.

That one line explains a table of moduli. Stiff, directional covalent bonds give several hundred gigapascals (alumina, silicon carbide) and diamond over 1000 GPa. Metallic bonds give about 70 GPa for aluminium, 120–130 for copper, 200–210 for steels and about 410 for tungsten; glasses sit near 70. Polymers have strong bonds along their chains but weak van der Waals forces between them, so most plastics have moduli of 1–4 GPa. Rubber, at a few megapascals, is a special case whose stiffness comes from the entropy of coiled chains (Polymers).

It also shows why stiffness resists processing. Heat treatment, cold work and grain refinement rearrange the defects in a metal, but nearly all its atoms still sit in the same wells, with the same curvature. To change EE appreciably you must change the bonds, that is, the material: a different alloy base, or stiff fibres in a composite.

DefinitionsH

The elastic constants

Definition— Elastic moduli

For an isotropic material, one whose properties are the same in every direction, four constants describe small elastic deformations, and any two determine the others:

  • Young's modulus EE: tensile stress over tensile strain, σ=Eε\sigma = E\varepsilon.
  • Shear modulus GG: shear stress over shear strain, τ=Gγ\tau = G\gamma.
  • Bulk modulus KK: pressure over the fractional change in volume, p=−K ΔV/Vp = -K\,\Delta V/V.
  • Poisson's ratio ν\nu: minus the lateral strain over the axial strain in a tensile test.

They are related by E=2G(1+ν)=3K(1−2ν)E = 2G(1+\nu) = 3K(1-2\nu).

Poisson's ratio is about 0.3 for most metals, 0.2 for many ceramics and glasses, nearly 0.5 for rubber, which keeps its volume, and nearly zero for cork, which is why a cork does not get fatter as it is pushed into a bottle. Designed lattices can even have a negative ratio (Metamaterials).

A single crystal is generally anisotropic: a cubic crystal needs three elastic constants, C11C_{11}, C12C_{12} and C44C_{44}, and its stiffness depends on direction. Iron is stiffest along the cube diagonal ⟨111⟩\langle 111 \rangle, at about 280 GPa, and least stiff along the cube edge ⟨100⟩\langle 100 \rangle, at about 130 GPa; tungsten is a rare exception, isotropic even as a single crystal. A polycrystal of randomly oriented grains averages over all directions, which is why an engineering table gives a single 200 GPa. Rolled sheet develops a texture, a preferred orientation of its grains, and becomes measurably anisotropic again, which matters to can makers and car-body press shops.

Note

Computed elastic tensors for thousands of crystals are free to look up in the Materials Project, calculated with density functional theory (see Density functional theory). Measured values for engineering alloys are in handbooks from societies such as ASM International. For a design, use measured data for the actual material and temperature.

ComputationC

Computing it: from a bond to a modulus and an ideal strength

The force curve has a maximum; pull harder and the bond comes apart. That maximum force divided by the area per bond is the ideal tensile strength of a perfect crystal. The program computes it for a Lennard-Jones bond, U=4ϵ[(s/r)12−(s/r)6]U = 4\epsilon[(s/r)^{12} - (s/r)^6] with ϵ=s=1\epsilon = s = 1, and compares it with E≈S0/r0E \approx S_0/r_0; for a Morse bond the ratio has the closed form 1/(4ar0)1/(4ar_0), where aa sets the width of the well. Last comes Yakov Frenkel's 1926 estimate for shear: sliding one plane of atoms rigidly over another needs about G/2πG/2\pi.

python
import numpy as np

# A Lennard-Jones bond in reduced units: well depth eps = 1, length scale s = 1
def U(r):
    return 4.0 * (r**-12 - r**-6)

r = np.linspace(1.0, 2.5, 1_500_001)
dUdr = np.gradient(U(r), r)              # the force needed to hold the atoms at separation r

i0 = np.argmin(np.abs(dUdr))             # equilibrium: no force
r0 = r[i0]
h = 1e-4
k = (U(r0 + h) - 2 * U(r0) + U(r0 - h)) / h**2   # bond stiffness U''(r0)
im = np.argmax(dUdr)                     # the largest restoring force the bond can supply
print(f"equilibrium spacing r0      = {r0:.4f}   (exact 2^(1/6) = {2**(1/6):.4f})")
print(f"bond stiffness U''(r0)      = {k:.2f}")
print(f"largest force at r          = {r[im]:.4f}  (stretch of {100*(r[im]/r0 - 1):.1f}%)")
print(f"largest force F_max         = {dUdr[im]:.3f}")
print(f"E ~ k/r0 = {k/r0:.1f};  sigma_max ~ F_max/r0^2 = {dUdr[im]/r0**2:.2f}")
print(f"ideal tensile strength / E  = {dUdr[im] / (k * r0):.3f}  = E/{k * r0 / dUdr[im]:.0f}")

# A Morse bond gives the ratio in closed form: 1 / (4 a r0), with a r0 typically 3 to 6
for ar0 in [3, 4, 5, 6]:
    print(f"Morse bond, a*r0 = {ar0}: ideal strength = E/{4 * ar0}")

# Frenkel's estimate of the ideal shear strength, for copper
G = 48e9                                 # Pa, shear modulus of polycrystalline copper (rounded)
tau_ideal = G / (2 * np.pi)
tau_real = 1e6                           # Pa: the order of the measured value for a pure crystal
print(f"\ncopper: ideal shear strength G/(2 pi) = {tau_ideal / 1e9:.1f} GPa")
print(f"measured critical resolved shear stress ~ 1 MPa: about {tau_ideal / tau_real:,.0f} times smaller")
Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

The bond gives way at a stretch of about 11%, and the ideal strength is about E/27E/27 for the Lennard-Jones bond and E/12E/12 to E/24E/24 for Morse bonds of typical width: whatever the details, a perfect crystal should be a tenth to a thirtieth as strong as it is stiff, about 7–17 GPa for steel.

Real steels yield at 0.2–2 GPa, and a pure copper crystal slips at a resolved shear stress of the order of 1 MPa, nearly eight thousand times below Frenkel's estimate. The puzzle has a single answer: crystals do not slip whole planes at once.

ExplanationH

Plastic flow: dislocations and slip

A dislocation, as Defects and dislocations showed, is the boundary line between a part of a slip plane that has slipped by one lattice vector, the Burgers vector b\mathbf b, and a part that has not. Moving it needs only a row of bonds at a time to change partners, so the stress required is tiny compared with slipping the whole plane at once.

ττslip planeextra half-planeafter the dislocationhas crossed the crystalbba step of one Burgersvector at each end
An edge dislocation (⊥) in a simple cubic crystal: above the slip plane there is one extra half-plane of atoms (thick line). Under the shear stress τ the dislocation glides one row of bonds at a time; when it has crossed the crystal, the top half has moved by one Burgers vector b (right).

Analogy, not a formal construction

Dragging a heavy carpet is hard work. Instead, make a ruck at one end and push it across: only a strip of carpet moves at any moment, and when the ruck reaches the far end the whole carpet has shifted by the size of the ruck. A dislocation is the ruck.

Slip systems. Dislocations glide most easily on the most densely packed (and most widely spaced) planes, in the most densely packed directions, where the Burgers vector is shortest. A plane with a direction in it is a slip system:

StructureExamplesSlip systemsConsequence
FCCAl, Cu, Ni, Ag, Au, austenitic steel{111}⟨110⟩\lbrace 111 \rbrace\langle 110 \rangle: 12Ductile, even at very low temperature
BCCFe (ferrite), W, Mo, Cr, Nb⟨111⟩\langle 111 \rangle directions on {110}\lbrace 110 \rbrace, {112}\lbrace 112 \rbrace and {123}\lbrace 123 \rbrace planesDuctile when warm; strength rises steeply on cooling
HCPMg, Zn, Ti, Zrbasal (0001)⟨112ˉ0⟩(0001)\langle 11\bar{2}0 \rangle: 3, plus harder prismatic and pyramidal systemsOften limited ductility; strongly anisotropic sheet

In 1928 Richard von Mises showed that a grain in a polycrystal needs five independent slip systems to change shape arbitrarily while staying in contact with its neighbours. FCC metals have them easily, which is why gold can be beaten into leaf a tenth of a micrometre thick; magnesium at room temperature has only its basal systems, which is why its sheet is formed warm.

Schmid's law. Only the shear stress resolved onto the slip plane, in the slip direction, drives a dislocation. If the tensile axis makes an angle ϕ\phi with the slip-plane normal and λ\lambda with the slip direction, the resolved shear stress is

τR=σcos⁡ϕcos⁡λ,\tau_R = \sigma \cos\phi \cos\lambda,

and the crystal yields when τR\tau_R on its most favourable system reaches the critical resolved shear stress τc\tau_c, a property of the material. The product m=cos⁡ϕcos⁡λm = \cos\phi\cos\lambda is the Schmid factor; it can never exceed 12\tfrac12, reached when both angles are 45°.

Why the stress is not zero. A dislocation must still push its core through the hills and valleys of the lattice, which needs the Peierls stress. In FCC metals the core spreads out on the smooth close-packed planes and the Peierls stress is tiny, so their strength comes almost entirely from obstacles. In BCC metals the screw-dislocation core is compact, the Peierls stress is large and is overcome partly by thermal vibration, so BCC metals grow much stronger, and eventually brittle, as they get colder: the root of the ductile-to-brittle transition in Fracture and toughness. In covalent crystals and most ceramics the Peierls stress is so high that at room temperature they fracture before they flow.

ComputationC

Computing it: Schmid factors in an FCC crystal

Which of the twelve FCC slip systems operates first depends on the direction of loading. The program lists the systems, computes every Schmid factor for four loading directions, and reports the stress at which a single crystal with τc=1\tau_c = 1 MPa would yield.

python
import itertools
import numpy as np

# The 12 slip systems of an FCC crystal: a {111} plane and a <110> direction lying in it
planes = [np.array(p) for p in [(1, 1, 1), (-1, 1, 1), (1, -1, 1), (1, 1, -1)]]
dirs = [np.array(v) for v in itertools.product((-1, 0, 1), repeat=3)
        if np.count_nonzero(v) == 2 and v[np.nonzero(v)[0][0]] > 0]   # one of each +/- pair
systems = [(n, d) for n in planes for d in dirs if n @ d == 0]
print("FCC slip systems:", len(systems))

def schmid(load, n, d):
    """cos(phi) cos(lambda) for tension along `load`, slip plane normal n, slip direction d."""
    t = np.asarray(load, float)
    return abs(t @ n) * abs(t @ d) / (np.linalg.norm(t) ** 2 * np.linalg.norm(n) * np.linalg.norm(d))

def fmt(v):
    return " ".join(str(int(x)) for x in v)

tau_c = 1.0                                   # MPa: an illustrative CRSS for a pure FCC metal
for load in [(0, 0, 1), (0, 1, 1), (1, 1, 1), (1, 2, 3)]:
    m = np.array([schmid(load, n, d) for n, d in systems])
    k = int(np.argmax(m))
    count = int(np.isclose(m, m[k]).sum())
    print(f"tension along [{fmt(load)}]: largest Schmid factor {m[k]:.3f} on {count} system(s), "
          f"e.g. ({fmt(systems[k][0])})[{fmt(systems[k][1])}]; the crystal yields at {tau_c / m[k]:.2f} MPa")

M = 3.06                                      # Taylor factor of a randomly oriented FCC polycrystal
print(f"\nrandomly oriented polycrystal: sigma_y = M * tau_c = {M * tau_c:.2f} MPa")
Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

Pulled along a cube edge, [001][001], eight systems share the largest Schmid factor, 1/6≈0.4081/\sqrt6 \approx 0.408; the crystal slips on several planes at once, the dislocations tangle, and it hardens quickly. Along [123][123] one system, (1ˉ11)[101](\bar1 1 1)[101], has the largest factor, 0.467, so the crystal yields sooner and at first glides easily on that system alone. Along [111][111] the best factor is only 0.272, and yielding needs 1.7 times the stress needed along [123][123].

A grain in a polycrystal cannot simply choose its easiest system, because it must deform compatibly with its neighbours. Taylor showed that a randomly oriented FCC polycrystal yields at about MτcM\tau_c, with the Taylor factor M≈3.06M \approx 3.06: the bridge from single crystals to the yield strength in an engineering table.

DefinitionsH

Strengthening: getting in the dislocations' way

If dislocations carry plastic flow, then everything that hinders their motion raises the yield strength. Metallurgists have four main ways to do it, and real alloys combine all four.

Definition— Grain-boundary strengthening and the Hall–Petch relation

A slip plane does not continue smoothly into a differently oriented grain, so dislocations pile up against the boundary, concentrating stress on it until slip starts next door. Smaller grains hold shorter pile-ups, so they need a larger applied stress. E. O. Hall (1951) and N. J. Petch (1953) found

σy=σ0+kyd,\sigma_y = \sigma_0 + \frac{k_y}{\sqrt d},

where dd is the grain size, σ0\sigma_0 the friction stress of the grain interiors and kyk_y the Hall–Petch coefficient: of order 0.1 MPa·m1/2^{1/2} or less for aluminium and copper, several times larger, about 0.6–0.7 MPa·m1/2^{1/2}, for ferritic steels.

Theorem— Why the exponent is one half (pile-up argument)

If dislocations from a source in a grain of size dd pile up against its boundary, and slip crosses the boundary when the stress on the leading dislocation reaches a critical value τ∗\tau^*, then the applied shear stress at yield satisfies τ−τ0∝τ∗Gb/d\tau - \tau_0 \propto \sqrt{\tau^* G b / d}, where τ0\tau_0 is the friction stress.

Proof

Advance the whole pile-up rigidly by δx\delta x. The effective stress τ−τ0\tau - \tau_0 does work n(τ−τ0)b δxn(\tau - \tau_0)b\,\delta x per unit length; the dislocations' mutual interactions are internal and cancel in pairs; only the obstacle resists, acting on the head dislocation. So τhead=n(τ−τ0)\tau_\text{head} = n(\tau - \tau_0). The equilibrium positions under mutual repulsion (Eshelby, Frank and Nabarro, 1951) give a pile-up of length L≈nGb/[π(1−ν)(τ−τ0)]L \approx nGb/[\pi(1-\nu)(\tau - \tau_0)] for edge dislocations. Setting L≈d/2L \approx d/2 gives n∝(τ−τ0)d/(Gb)n \propto (\tau - \tau_0)d/(Gb), and then τhead=τ∗\tau_\text{head} = \tau^* gives (τ−τ0)2∝τ∗Gb/d(\tau - \tau_0)^2 \propto \tau^* Gb/d. The Taylor factor converts these shear stresses into σ0\sigma_0 and kyk_y.

∎

Other models also give d−1/2d^{-1/2}, so the exponent alone does not prove the mechanism. Grain refinement has a special virtue: it is the only one of the four mechanisms that usually makes a steel tougher as well as stronger, because fine grains also stop cleavage cracks. Modern structural steels exploit it through thermomechanically controlled rolling: small additions of niobium, titanium or vanadium form particles that stop austenite grains growing during hot rolling, and the transformation on cooling (Processing and heat treatment) gives ferrite grains of a few micrometres.

Definition— Solid-solution strengthening

Solute atoms of a different size or stiffness from the host distort the lattice around them and interact with the stress fields of dislocations, pinning them. Strengthening rises with concentration cc, roughly as c1/2c^{1/2} (Fleischer's model, for dilute, strong obstacles) or c2/3c^{2/3} (Labusch's, for many weak ones).

Brass (copper with zinc) and aluminium–magnesium alloys rely on it. Interstitial carbon and nitrogen in iron distort the lattice asymmetrically and strengthen far more per atom; they also gather around dislocations in Cottrell atmospheres and pin them, which gives mild steel its sharp yield point. In the high-entropy alloys studied intensively since 2004, every atom is in a sense a solute.

Definition— Precipitation (age) hardening

Fine particles of a second phase obstruct dislocations. A dislocation can either cut through a small, coherent particle, paying for the new interface and any disorder it creates, or, if the particles are larger and stronger, bow between them and leave a loop around each one (the Orowan mechanism), which needs a shear stress of about Gb/LGb/L for particles a distance LL apart.

It was discovered by accident: around 1906 Alfred Wilm, near Berlin, found that a quenched aluminium–copper–magnesium alloy grew harder over several days at room temperature, though his microscopes showed no change. The particles responsible, a few atoms thick, were detected only in 1938, by X-ray diffraction, independently by André Guinier and George Preston. Wilm's duralumin built airships and metal aircraft; today the strongest aluminium alloys (2xxx, 6xxx, 7xxx), maraging steels and the nickel superalloys of jet engines, more than half of whose volume can be ordered γ′\gamma' particles, are all precipitation-hardened by the heat treatments of Processing and heat treatment.

Definition— Work (strain) hardening

Plastic deformation multiplies dislocations, from about 101010^{10}–101210^{12} m−2^{-2} (metres of line per cubic metre) in an annealed metal to 101510^{15} m−2^{-2} or more after heavy cold work. They intersect, form locks and tangles, and obstruct one another. The flow stress follows the Taylor relation

σ=σ0+MαGbρ,\sigma = \sigma_0 + M\alpha G b\sqrt{\rho},

with α\alpha a constant of about 0.2–0.5 and ρ\rho the dislocation density.

Work hardening is why a coppersmith's hammered bowl is harder to dent than the sheet it came from and why cold-drawn wire is strong. Annealing undoes it, as recovery and recrystallisation sweep the dislocations away.

ComputationC

Computing it: a strength budget

The first program estimates work hardening in copper and the Hall–Petch strength of a steel over four decades of grain size, with illustrative constants of the size reported for ferritic steels.

python
import numpy as np

# 1. Work hardening: the Taylor relation  sigma = M * alpha * G * b * sqrt(rho), for copper
M, alpha, G, b = 3.06, 0.3, 48e9, 0.256e-9      # Taylor factor, constant, Pa, m
print("dislocation density   extra yield strength")
for rho in [1e10, 1e12, 1e13, 1e14, 1e15]:      # metres of line per cubic metre
    ds = M * alpha * G * b * np.sqrt(rho) / 1e6
    spacing = 1 / np.sqrt(rho) * 1e9
    print(f"  {rho:7.0e} m^-2      {ds:6.1f} MPa   (dislocations ~{spacing:7.0f} nm apart)")

# 2. Grain refinement: the Hall-Petch relation  sigma_y = sigma_0 + k / sqrt(d)
sigma0, k = 100.0, 0.6                           # MPa and MPa m^0.5: illustrative, ferritic-steel-like
print("\ngrain size    sigma_y")
for d_um in [200, 50, 20, 10, 5, 2, 1, 0.5, 0.1]:
    s = sigma0 + k / np.sqrt(d_um * 1e-6)
    print(f"{d_um:7.1f} um  {s:6.0f} MPa  " + "#" * int(s / 40))
d_needed = (k / (500 - sigma0)) ** 2
print(f"\ngrain size for sigma_y = 500 MPa: {d_needed * 1e6:.2f} um")
Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

Annealed copper, at about 101210^{12} m−2^{-2}, gains only 11 MPa from its dislocations; heavy cold work, at 101510^{15}, adds about 360 MPa, roughly what hard-drawn copper wire achieves, and each tenfold gain costs a hundredfold more dislocations. In the steel, refining the grains from 200 µm to 5 µm raises the yield strength from about 140 to 370 MPa, and 500 MPa needs grains of about 2.25 µm. The 2000 MPa of the last line would need 0.1 µm grains, which only severe plastic deformation can make.

The second program shows why precipitation hardening has an optimum. During ageing the precipitates grow at nearly constant volume fraction ff, so their spacing grows with their radius rr. Cutting gets harder as they grow (as fr\sqrt{fr}, in theory), bowing gets easier, and the dislocation takes the easier path. We set the one free constant so that the crossover falls at 5 nm, a typical particle size in peak-aged aluminium alloys.

python
import numpy as np

# Precipitation hardening in an aluminium alloy: cut the particles, or bow round them?
G, b, M = 26e9, 0.286e-9, 3.06          # Pa, m, Taylor factor (aluminium, rounded)
f = 0.02                                # volume fraction of precipitates, held fixed during ageing

r = np.array([0.5, 1, 2, 3, 4, 5, 6, 8, 12, 20, 40]) * 1e-9            # particle radius, m
L = r * np.sqrt(2 * np.pi / (3 * f))    # mean spacing of the particles cut by a slip plane
tau_bow = G * b / L                     # Orowan: bow the dislocation between particles
r_star = 5e-9                           # put the crossover here, typical of peak-aged alloys
kappa = (G * b / (r_star * np.sqrt(2 * np.pi / (3 * f)))) / np.sqrt(f * r_star / b)
tau_cut = kappa * np.sqrt(f * r / b)    # shearing: grows as (f r)^(1/2)
tau = np.minimum(tau_cut, tau_bow)      # the dislocation takes the easier path

print(" radius   spacing   cut    bow  (MPa)   extra sigma_y")
for ri, Li, tc, tb, t in zip(r, L, tau_cut, tau_bow, tau):
    way = "both" if np.isclose(tc, tb) else ("cut" if tc < tb else "bow")
    s = M * t / 1e6
    print(f"{ri*1e9:5.1f} nm {Li*1e9:6.0f} nm {tc/1e6:5.0f} {tb/1e6:6.0f}         {s:5.0f} MPa {way:4} " + "#" * int(s / 15))
Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

Small particles are cut, and strength rises as they grow (under-aged); large ones are bypassed, and strength falls as they coarsen (over-aged). The peak, here about 445 MPa, is where the mechanism changes. That is why an alloy is aged for a set time at a set temperature, and why a part that overheats, or the heat-affected zone of a weld, loses strength for good. Real alloys combine several mechanisms, and contributions do not always simply add (obstacles of similar strength and spacing combine more like the root of the sum of squares), but a budget like this one, checked against a few tensile tests, is usually close enough to tell a promising composition from a hopeless one.

ExplanationH

Hardness tests

Hardness is resistance to indentation: press a hard indenter into a surface with a known force and measure the dent. It is the quickest mechanical test, needs only a small flat area and barely marks the part, so it checks welds, heat treatments and production parts.

  • Brinell (Johan August Brinell, Sweden, 1900): a hard ball, often 10 mm across, under a large load; the dent's diameter gives the hardness, averaged over coarse microstructures such as castings.
  • Vickers (Vickers Ltd, Britain, 1920s): a square diamond pyramid with 136° between opposite faces; HV=1.8544 F/d2HV = 1.8544\,F/d^2, with FF in kilograms-force and dd the mean diagonal in millimetres. The dents are geometrically similar at every size, so one scale works from lead to ceramics, and small loads (microhardness) can probe single phases or traverse a weld.
  • Rockwell (Hugh and Stanley Rockwell, United States, patented 1919): the machine reads the depth of penetration under a major load after a preload, with no microscope; HRC, with a diamond cone, is standard for hardened steels.
  • Instrumented (nano)indentation records load and depth continuously; the method of Oliver and Pharr (1992) extracts hardness and elastic modulus from the elastic unloading curve at depths of tens of nanometres, for thin films and individual phases.

They are standardised in ISO 6506, 6507, 6508 and 14577 and ASTM E10, E18, E92 and E384.

What hardness measures. Under the indenter the metal flows plastically in a constrained region, and David Tabor showed around 1950 that the mean contact pressure is about three times the flow stress: H≈3σH \approx 3\sigma with HH in pressure units (multiply a Vickers number by 9.81 for MPa). That is why hardness correlates with tensile strength, and why empirical conversion tables are published (ISO 18265, ASTM E140); they differ between alloy families and are a check, not a substitute for a tensile test.

Example— A Vickers test on a structural steel

A 10 kgf load leaves a dent with diagonals of 0.312 mm and 0.308 mm. The mean diagonal is 0.310 mm, so HV=1.8544×10/0.3102=193HV = 1.8544 \times 10 / 0.310^2 = 193, written 193 HV10. In pressure units that is 193×9.81≈1890193 \times 9.81 \approx 1890 MPa, and Tabor's factor of three suggests a flow stress of roughly 630 MPa. Because the indentation itself strains the metal by several per cent, that figure compares best with the tensile strength of a work-hardening steel, which for this hardness is typically 600–650 MPa, rather than its yield strength.

ExplanationH

In professional practice

Alloy design. New alloys are designed by combining thermodynamic calculations of which phases form (the CALPHAD method, with commercial databases and open-source tools such as pycalphad), kinetic models of precipitate growth and strength models like the budget above. This integrated computational materials engineering (ICME) has shortened the development of some aerospace steels and superalloys from a decade or more to a few years.

Simulation. Molecular dynamics, with codes such as LAMMPS and potentials from the NIST Interatomic Potentials Repository, follows single dislocations through obstacles; as of 2026, machine-learned interatomic potentials trained on quantum calculations describe dislocation cores in alloys far better than classical ones. Discrete dislocation dynamics follows millions of segments, and crystal plasticity finite-element models build Schmid's law into every grain to predict textures and the scatter of strength (Multiscale modelling).

Data and testing. Elastic constants are among the most reliably computed properties, and the Materials Project's computed elastic tensors are widely used for screening. Strength depends on processing as much as composition, so measured data for the actual condition remain essential: tensile tests (ISO 6892-1, ASTM E8/E8M) for design values, hardness for production, and, since 2004, compression of micrometre-sized pillars cut with a focused ion beam to measure single grains and phases.

History & global context

A short history

Smiths hardened and softened metals by hammering, quenching and tempering for millennia before anyone knew why. The puzzle of strength came into focus when Frenkel's 1926 estimate showed crystals to be thousands of times weaker than their bonds. In 1934 Egon Orowan, Michael Polanyi and G. I. Taylor independently proposed the dislocation, building on Vito Volterra's elastic theory of such defects (1907). Dislocations stayed a theory until 1956, when Peter Hirsch, Robert Horne and Michael Whelan in Cambridge, and Walter Bollmann in Geneva, saw them moving in thin metal foils in the transmission electron microscope. With Cottrell's explanation of the yield point (1949) and the grain-size relation of Hall and Petch, metallurgy became a science of microstructure.

Common misconceptionsH

Common misconceptions

Misconception

"A stronger steel is stiffer." A quenched-and-tempered alloy steel with ten times the yield strength of a mild steel has the same Young's modulus to within a few per cent, so under the same load it deflects the same amount. A designer who must reduce deflection changes the shape (a deeper section) or the material family, not the heat treatment.

Misconception

"Yield strength is a fixed property of a material." It depends on grain size, cold work, ageing, temperature and strain rate: the aluminium alloy 6061 yields at about 55 MPa annealed and about 275 MPa in the peak-aged T6 condition.

Misconception

"Smaller grains always mean a stronger metal." Below roughly 10–30 nm, grains are too small to hold pile-ups, deformation shifts to the grain boundaries themselves, and the strength levels off or falls. Fine grains also creep faster when hot (Fatigue, creep and corrosion).

Misconception

"Plastic deformation breaks bonds." In a metal, dislocation glide only switches bonds from one neighbour to the next; the crystal stays a crystal, and its density barely changes. That is why a metal can be bent, rolled and drawn by enormous amounts without coming apart, and why its stiffness is unchanged afterwards.

Exercises

Exercises

1. Elasticity, plasticity and strengthening

Warm-upHuman

A steel bracket bends too much under its working load, although it never yields. Its shape cannot change. Which of these would noticeably reduce the deflection?

Choose one

2. Elasticity, plasticity and strengthening

CoreHuman

An FCC single crystal is pulled in tension along the cube edge [001][001]. How many of its twelve {111}⟨110⟩\lbrace 111 \rbrace\langle 110 \rangle slip systems share the largest Schmid factor?

3. Elasticity, plasticity and strengthening

Warm-upHuman

A steel obeys σy=σ0+ky/d\sigma_y = \sigma_0 + k_y/\sqrt d with σ0=150\sigma_0 = 150 MPa and ky=0.5k_y = 0.5 MPa·m1/2^{1/2}. What is its yield strength, in MPa, when the grain size is d=25d = 25 µm?

4. Elasticity, plasticity and strengthening

CoreHuman

A Vickers test with a load of 30 kgf leaves a dent whose two diagonals measure 0.418 mm and 0.422 mm. Using HV=1.8544 F/d2HV = 1.8544\,F/d^2 (FF in kgf, dd the mean diagonal in mm), what is the Vickers hardness, rounded to the nearest whole number?

5. Elasticity, plasticity and strengthening

CoreHuman

A shipbuilder needs a plate steel with a higher yield strength that is also tougher at low temperature. Which strengthening mechanism is the best choice?

Choose one

6. Elasticity, plasticity and strengthening

CoreComputational

Write fit_hall_petch(d_um, sigma_mpa), which takes a list of grain sizes in micrometres and the measured yield strengths in MPa, fits σy=σ0+kyd−1/2\sigma_y = \sigma_0 + k_y d^{-1/2} by least squares with dd in metres, and returns the tuple (sigma0, k) with σ0\sigma_0 in MPa and kyk_y in MPa·m1/2^{1/2}.

Try it on these yield strengths of a ferritic steel at grain sizes 100, 40, 16, 6.25, 4 and 1 µm: 178, 205, 254, 336, 390 and 668 MPa (made-up practice data, not a real test series).

Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

7. Elasticity, plasticity and strengthening

StretchComputational

Write max_schmid(load), which returns the largest Schmid factor cos⁡ϕcos⁡λ\cos\phi\cos\lambda among the twelve {111}⟨110⟩\lbrace 111 \rbrace\langle 110 \rangle slip systems of an FCC crystal pulled along the direction load (a list of three numbers, not necessarily a unit vector, of either sign). For example, max_schmid([0, 0, 1]) is 1/6≈0.4081/\sqrt6 \approx 0.408.

Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

Further reading and resources