Fibres in a matrix
About half of the structure of a Boeing 787 or an Airbus A350, by weight, is carbon-fibre composite. The longest wind-turbine blades in service exceed 100 m, almost all glass fibre in polymer resin, and the hull of most small boats, the frame of a racing bicycle, a tennis racket, a pole-vaulting pole and the circuit board in a phone are composites too. So is wood, of cellulose fibres in a lignin matrix, and so is bone.
The idea is old. Mud bricks were reinforced with straw; the composite bows of Central Asian horse peoples laminated horn, wood and sinew so that each worked where it was best; reinforced concrete put steel bars where concrete is weak. The modern era began with glass fibre in the 1930s and 1940s, and carbon fibre in the 1960s, developed from polyacrylonitrile (PAN) by Akio Shindo in Japan and by William Watt, Leslie Phillips and William Johnson at the Royal Aircraft Establishment in Britain.
A composite combines two or more materials, separate and distinct at the microscopic scale, so that the whole has properties neither has alone. This lesson concentrates on the commonest kind, fibre-reinforced polymers: why fibres, how to calculate the stiffness of the result, why it depends on direction, how such parts are made and used, and the hard question of what happens to them when they are retired.
Why fibres, and what the matrix does
In ceramics and glasses you saw that brittle solids fail from their worst flaw, and that a smaller volume holds fewer bad flaws. Around 1920 A. A. Griffith drew glass into fibres and found that the thinner they were, the stronger: his finest fibres were more than ten times as strong as ordinary glass rods (Griffith). A fibre a few micrometres across can contain only tiny flaws, and when the drawing or spinning process lines up its molecules or crystals along the axis (as in carbon, aramid and polyethylene fibres), the strongest bonds point along the load. So the strongest and stiffest materials we can make come as fibres, and a fibre on its own is useless for a beam: it buckles in compression and has no stiffness across.
The matrix, usually a polymer, makes fibres into a material. It holds them in place and in line, transfers load into and between them by shear at the interface, protects them from abrasion and moisture, and stops a crack in one fibre from running into the next. The interface must be strong enough to carry load but, in many composites, weak enough to let a crack turn and run along the fibre instead of straight through the next one; the fibres then pull out, absorbing energy. That is why a composite of two brittle materials, glass and a glassy polymer, can be tough. Fibres are coated with a sizing during manufacture to control this bond.
Fibres and matrices
Typical properties of the main reinforcing fibres, from producers' data sheets (natural fibres vary widely with plant and processing):
| Fibre | Young's modulus (GPa) | Tensile strength (GPa) | Density (g/cm³) | Notes |
|---|---|---|---|---|
| E-glass | 72–73 | 2–3.5 | 2.55–2.6 | cheap, insulating; boats, blades, circuit boards |
| carbon, standard modulus (PAN) | about 230 | 3.5–5 | 1.76–1.80 | most aerospace and sport |
| carbon, intermediate/high modulus | 290–600 (pitch-based, to 900) | 3–6 | 1.8–1.9 | stiffness-critical structures, satellites |
| aramid | 70–140 | about 3 | 1.44 | tough, impact- and cut-resistant; weak in compression |
| ultra-high-molecular-weight polyethylene | about 100 | about 3.5 | 0.97 | lightest; ropes, protective equipment; melts near 150 °C |
| basalt | about 90 | about 3 | 2.7 | from volcanic rock |
| flax | 50–70 | 0.5–1.5 | 1.45–1.5 | renewable, damps vibration |
The matrices are mostly polymers. Thermosets — epoxies for aerospace and sport, unsaturated polyesters and vinyl esters for boats, tanks and pipes — are cheap to process as low-viscosity liquids that soak into the fibres and then cure. Thermoplastics such as polypropylene, polyamides and the high-temperature PEEK can be melted, welded and, in principle, recycled, but their melts are very viscous and harder to work into the fibres. Beyond polymers, metal-matrix composites (aluminium reinforced with silicon carbide particles, for brake discs and electronic heat sinks) and ceramic-matrix composites (silicon carbide fibres in silicon carbide, now flying in the hot sections of jet engines) push the idea to high temperature.
Definition— lamina and laminate
A lamina or ply is a thin layer of fibres all running in one direction (unidirectional) or woven in two, embedded in matrix. A laminate is a stack of plies bonded together, each laid at a chosen angle; its lay-up lists the angles, for example , where means the stack is repeated in mirror order to make it symmetric about its mid-plane.
The rule of mixtures and its bounds
How stiff is a composite? Start with two simple arrangements of fibre and matrix in layers.
Along the layers, both phases share one strain ; the force is the sum of the forces in each, so , and , where and are the volume fractions. Across the layers, both carry the same stress; the extensions add, so . These are the rule of mixtures and the inverse rule of mixtures. They are more than two special cases.
Theorem— Voigt and Reuss bounds
For any arrangement of two linear elastic phases with volume fractions and , the effective Young's modulus in any direction lies between the Reuss and Voigt averages:
Proof
We sketch the argument of Hill (1952), for a specimen stretched along one axis. The true deformation minimises the potential energy among all displacement fields consistent with the imposed stretch, so the stored energy is no larger than that of any trial field. Try the uniform strain everywhere: its energy is per unit volume (taking each phase's stiffness for that stress state as its Young's modulus, which is exact for the slab and, for full three-dimensional stress states, gives Hill's slightly more general form). So , the upper bound. Dually, the true stress field minimises the complementary energy among all stress fields in equilibrium with the applied load; trying the uniform stress gives , the lower bound. The slab models attain each bound, so neither can be improved without more information about the geometry.
∎
Fibres are not slabs, so for the transverse modulus engineers use semi-empirical fits between the bounds, such as the Halpin–Tsai equation (Halpin and Kardos), or tighter rigorous bounds (Hashin and Shtrikman), or a finite-element model of a fibre arrangement. Here are the bounds and the Halpin–Tsai estimate for glass fibre in epoxy.
import numpy as np
# E-glass fibres in epoxy (both roughly isotropic). Moduli in GPa, densities in g/cm^3.
Ef, Em = 72.0, 3.5
rho_f, rho_m = 2.55, 1.20
def voigt(Vf): # equal strain: load along the fibres
return Vf * Ef + (1 - Vf) * Em
def reuss(Vf): # equal stress: load across the layers
return 1 / (Vf / Ef + (1 - Vf) / Em)
def halpin_tsai(Vf, xi=2.0): # semi-empirical transverse modulus, circular fibres
eta = (Ef / Em - 1) / (Ef / Em + xi)
return Em * (1 + xi * eta * Vf) / (1 - eta * Vf)
print(" Vf E1 Voigt E2 Reuss E2 Halpin-Tsai density E1/rho")
for Vf in (0.0, 0.2, 0.4, 0.5, 0.6, 0.7):
rho = Vf * rho_f + (1 - Vf) * rho_m
print(f"{Vf:4.1f} {voigt(Vf):9.1f} {reuss(Vf):10.1f} {halpin_tsai(Vf):14.1f} {rho:10.2f} {voigt(Vf) / rho:8.1f}")
# Carbon fibre (axial modulus 230 GPa, density 1.76) at Vf = 0.6, for comparison:
Vf = 0.6
E1_cf = Vf * 230 + (1 - Vf) * Em
rho_cf = Vf * 1.76 + (1 - Vf) * rho_m
print(f"\ncarbon/epoxy, Vf = 0.6: E1 = {E1_cf:.0f} GPa, density {rho_cf:.2f}, E1/rho = {E1_cf / rho_cf:.0f}"
f" (steel and aluminium: about 26)")
# Critical fibre length (Kelly-Tyson): a fibre must be at least l_c long to be loaded to its
# strength sigma_f through an interface that can carry a shear stress tau.
d, sigma_f, tau = 7e-6, 3.5e9, 50e6
print(f"critical length for a 7 um carbon fibre: {sigma_f * d / (2 * tau) * 1e3:.2f} mm")At a typical fibre fraction of 0.6, glass/epoxy is 44.6 GPa along the fibres but only 8–15 GPa across them; the true transverse value, measured, is usually close to the Halpin–Tsai figure. Carbon fibre in the same matrix gives 139 GPa along the fibres, at a density of 1.54 g/cm³, a specific stiffness of 91 against about 26 for every common structural metal (metals and alloys). That number, in one direction, is the whole case for carbon fibre.
Strength follows a similar rule along the fibres, , where is the matrix stress when the fibres break, but only if the fibres are long enough to be loaded fully. Load enters a fibre by shear at the interface, building up from zero at its ends. The critical length , about a quarter of a millimetre for a carbon fibre 7 µm across with an interface shear strength of 50 MPa, is the shortest fibre that can be loaded to its breaking stress. Chopped fibres in injection-moulded parts, often under a millimetre long after moulding, are only a few times and use their strength less efficiently than continuous ones.
Fibre orientation and laminates
A unidirectional ply is a strongly anisotropic material. Its stiffness in a direction at angle to the fibres follows from rotating its elastic constants:
where and are the moduli along and across the fibres, the in-plane shear modulus and the major Poisson's ratio. A part loaded in several directions therefore needs plies in several directions. Classical laminate theory adds up the stiffness of each ply, rotated to its angle and weighted by its thickness, into a matrix relating in-plane forces to strains (and matrices and for bending, which we leave out by keeping laminates symmetric). The program builds for three lay-ups of carbon/epoxy and reads off the modulus in every direction.
import numpy as np
# A unidirectional carbon/epoxy ply. E1 from the rule of mixtures above; E2, G12 and nu12 are
# typical measured values (the transverse properties cannot be computed from the fibre's axial
# modulus, because carbon fibres are themselves anisotropic).
E1, E2, G12, nu12 = 139.0, 9.0, 5.0, 0.30 # GPa
nu21 = nu12 * E2 / E1
def Q_bar(theta_deg):
"""Reduced stiffness matrix of a ply rotated by theta, in the laminate axes (GPa)."""
d = 1 - nu12 * nu21
Q = np.array([[E1 / d, nu12 * E2 / d, 0], [nu12 * E2 / d, E2 / d, 0], [0, 0, G12]])
c, s = np.cos(np.radians(theta_deg)), np.sin(np.radians(theta_deg))
T = np.array([[c * c, s * s, 2 * c * s], [s * s, c * c, -2 * c * s], [-c * s, c * s, c * c - s * s]])
R = np.diag([1, 1, 2]) # engineering vs tensor shear strain
return np.linalg.inv(T) @ Q @ R @ T @ np.linalg.inv(R)
def in_plane_modulus(layup, phi):
"""Young's modulus of a symmetric laminate (equal ply thicknesses) in direction phi."""
A = sum(Q_bar(theta - phi) for theta in layup) / len(layup) # A matrix / thickness
return 1 / np.linalg.inv(A)[0, 0]
layups = {
"unidirectional [0]8": [0] * 8,
"cross-ply [0/90]2s": [0, 90, 0, 90, 90, 0, 90, 0],
"quasi-isotropic [0/45/-45/90]s": [0, 45, -45, 90, 90, -45, 45, 0],
}
angles = (0, 15, 30, 45, 60, 90)
print("modulus (GPa) along a direction phi from the 0-degree fibres")
print(f"{'laminate':32}" + "".join(f"{a:>7}" for a in angles))
for name, layup in layups.items():
print(f"{name:32}" + "".join(f"{in_plane_modulus(layup, a):7.1f}" for a in angles))
print(f"{'aluminium alloy (isotropic)':32}" + "".join(f"{70.0:7.1f}" for a in angles))The unidirectional laminate is 139 GPa along its fibres but already down to 52.7 GPa at 15° and 12.7 GPa at 45°, not far above its transverse value of 9 GPa. The cross-ply is stiff at 0° and 90° (74.3 GPa) but weak in shear, so soft at 45°. The quasi-isotropic lay-up, with equal plies at 0°, ±45° and 90°, has the same in-plane modulus, 53.4 GPa, in every direction: less than aluminium's 70 GPa, but at 57 % of the density, so it is about a third stiffer per kilogram. Real aircraft laminates are tailored between these extremes, with more 0° plies where the main load runs and ±45° plies to carry shear and torsion.
Laminates fail in ways metals do not: matrix cracks in plies loaded across their fibres, delamination between plies, fibre breakage, and in compression, kinking of fibres that the matrix can no longer hold straight. Their weakest property is usually strength through the thickness, where only the matrix carries load.
Carbon, glass and natural fibres at work
Carbon-fibre composites are chosen where stiffness and strength per kilogram justify their cost. In aircraft, wing skins, spars and fuselage barrels of carbon/epoxy save weight and do not fatigue or corrode as aluminium does. In racing and supercars, the moulded passenger cell is a carbon monocoque that protects the driver in a crash by crushing progressively; carbon-fibre road cars are still rare, though one German small electric car of 2013–2022 had a carbon passenger cell. Carbon-fibre filament-wound tanks with polymer liners hold hydrogen for fuel-cell vehicles at 70 MPa. And sport is carbon's showcase: bicycle frames under a kilogram, racket and golf-club shafts, rowing shells, hockey sticks and prosthetic running blades.
Glass-fibre composites are the volume business, cheaper and tougher in impact if less stiff. Boat hulls, storage tanks, pipes, car and lorry panels, bathroom fittings and the FR-4 glass/epoxy board of almost every printed circuit (see printed circuit boards in Lumenary EE) are glass. So are most wind-turbine blades, made of glass/epoxy skins over balsa or foam cores, with carbon fibre added in the spar caps of the longest ones, where stiffness keeps the tip from striking the tower (see wind and water turbines in Lumenary ME).
Natural-fibre composites, of flax, hemp, jute, kenaf or sisal in polymers, are used in car door panels and parcel shelves, in sports equipment that benefits from their vibration damping, and in some boat and building panels. Their specific stiffness matches or exceeds E-glass's, they need far less energy to make, and they grow in many countries; their drawbacks are variability, moisture absorption and a low processing-temperature limit before the fibres degrade. Aramid fibres, tough and cut-resistant, go into protective gloves, helmets, cables and the impact-resistant layers of hybrid laminates.
Making composite parts
A composite part and its material are made at the same time, so the process sets the quality. The main routes, roughly from cheapest to most demanding:
- Hand lay-up and spray-up: fabric or chopped glass wetted with resin in an open mould; boats and tanks.
- Compression moulding of sheet moulding compound (chopped glass in polyester paste), and injection moulding of short-fibre thermoplastics: car parts in large numbers.
- Pultrusion: fibres pulled through a resin bath and a heated die, making continuous profiles, rods and spar caps. Filament winding: fibres wound around a rotating mandrel, for pipes and pressure vessels.
- Vacuum infusion and resin transfer moulding: dry fibres placed in a closed mould or under a vacuum bag, then resin drawn or pumped in; wind-turbine blades, boats, car parts.
- Prepreg and autoclave: fibres pre-impregnated with partly cured epoxy, laid ply by ply by hand or by automated fibre placement robots, vacuum-bagged and cured at around 180 °C under several bar of pressure in an autoclave; aircraft structures, with void contents below about 1 %. Out-of-autoclave prepregs and welded thermoplastic parts are reducing the cost.
Infusion is a race between flow and cure. Resin flows through a dry fibre stack as through a porous rock, following Darcy's law, and the time for the flow front to travel a distance grows as :
# Vacuum infusion: how long does resin take to soak through a dry fibre stack?
# Darcy's law for flow from a line inlet: the flow front reaches distance L after
# t = mu * phi * L**2 / (2 * K * dP)
mu = 0.25 # resin viscosity, Pa s (a low-viscosity infusion epoxy at room temperature)
phi = 0.5 # porosity of the dry stack (fibre volume fraction 0.5)
dP = 0.9e5 # driving pressure, Pa: atmosphere pushing resin into a vacuum bag
gel_time = 3600 # s: the resin starts to cure and stops flowing after about an hour
for K, label in [(2e-10, "fabric alone"), (2e-9, "with a flow mesh on top")]:
print(f"permeability {K:.0e} m^2 ({label})")
for L in (0.25, 0.5, 1.0, 2.0):
t = mu * phi * L**2 / (2 * K * dP)
flag = "" if t < gel_time else " <- gels before it fills"
print(f" flow length {L:4.2f} m: {t / 60:7.1f} min{flag}")With these representative values, resin crosses a metre of fabric in about an hour, close to its gel time, and two metres would take four hours. A highly permeable flow mesh on top of the stack cuts the times tenfold, and a long blade is fed from many inlets along its length so that no flow path is longer than a metre or two. Process engineers simulate the filling of each new mould to place inlets and vents where dry spots cannot form.
End of life
The cross-linked thermosets that make composites easy to process also make them hard to recycle: the matrix cannot be remelted, and fibres and matrix are intimately mixed. Wind energy has made the problem visible. Blades last twenty to twenty-five years, and the first large generation is now being retired; one widely cited estimate projected 43 million tonnes of cumulative blade waste worldwide by 2050 (Liu and Barlow). Several European countries have banned landfilling composite waste, and the European wind industry called for a Europe-wide ban on landfilling blades from 2025 (WindEurope).
The routes available, as of 2026:
- Mechanical recycling: shredding and grinding into filler or short fibres for lower-grade products. Simple, but most of the value is lost.
- Cement co-processing: shredded glass-fibre waste is fed to cement kilns, where the resin burns as fuel and the glass's silica, alumina and calcium become raw material for clinker. It is operating at industrial scale in Germany and elsewhere.
- Pyrolysis and solvolysis: heat or solvents decompose the resin and release the fibres. For carbon fibre, which is costly to make, recovered fibre retains most of its stiffness but comes out short and tangled; it is sold into non-woven mats and moulding compounds.
- Designing for recycling: thermoplastic matrices, including a liquid acrylic resin that can be infused like a thermoset and later dissolved or depolymerised (a 62 m prototype blade was made with it in 2022), and cleavable epoxies whose cross-links can be broken in a mild acid bath, used in commercial recyclable blades since the early 2020s. Vitrimers (polymers) are another route still in research.
Lifetime matters too: a composite that saves fuel for thirty years in an aircraft may have a far better life-cycle balance than its end-of-life difficulty suggests, while a composite part replacing an easily recycled metal in a short-lived product may not. Those comparisons are taken up in sustainable materials.
How composite structures are designed and certified
Composite structures are designed with laminate analysis and finite-element models (see finite element method in Lumenary ME), using failure criteria applied ply by ply, and allowables derived statistically from coupon tests, in the building-block approach described above; the CMH-17 handbook gathers the data and practice used across the aerospace industry. Aviation authorities such as the FAA and EASA certify composite airframes against requirements for damage tolerance, environmental effects, lightning strike (carbon is conductive but far less so than aluminium, so composite skins carry metal meshes) and fire. Parts are inspected by ultrasonic scanning, thermography and X-ray computed tomography, and in service by scheduled non-destructive inspection.
Research in 2026 includes automated fibre placement of thermoplastic tapes welded in place without an autoclave, multifunctional composites that store energy or carry sensors, bio-based resins and fibres, digital twins that follow each part from layup to retirement, and ceramic-matrix composites for hotter, more efficient turbines.
Common misconceptions
Misconception
"Carbon fibre is stronger than steel." Along its fibres, per kilogram, yes, by a wide margin. Across the fibres, or through the thickness, a carbon laminate is only as strong as its epoxy matrix, tens of megapascals. Composite design is the art of loading fibres along their length.
Misconception
"The rule of mixtures gives a composite's stiffness." It gives the stiffness along aligned continuous fibres, and an upper bound in every other direction. Across the fibres the stiffness is much lower, near the inverse rule of mixtures.
Misconception
"Composites don't corrode or fatigue." They do not rust, but they absorb moisture, which lowers the matrix's glass-transition temperature, and they fatigue by gradual matrix cracking and delamination. Carbon fibre in contact with aluminium sets up galvanic corrosion of the aluminium, which is why carbon airframes use titanium fittings and fasteners and insulating layers at joints.
Misconception
"A lighter composite part is always the greener choice." It depends on how long it is used and what saving it brings. The energy to make carbon fibre is high and the matrix is hard to recycle, so the case is strongest for long-lived parts in things that move.
Exercises
1. Composites
Warm-upHumanA unidirectional carbon/epoxy ply has a fibre volume fraction of 0.55. The fibres' axial modulus is 230 GPa and the epoxy's modulus is 3.5 GPa. Using the rule of mixtures, what is the ply's Young's modulus along the fibres, in GPa? Give the nearest whole number.
2. Composites
CoreHumanA designer estimates the stiffness of a unidirectional glass/epoxy ply loaded perpendicular to its fibres using , and gets 44.6 GPa at . What is wrong?
3. Composites
CoreHumanShort glass fibres 15 µm in diameter, with a tensile strength of 2.0 GPa, are moulded into a polyamide. The fibre–matrix interface can carry a shear stress of 30 MPa. What is the critical fibre length, in micrometres?
4. Composites
CoreHumanWhy are most structural laminates both symmetric about their mid-plane (the subscript in ) and balanced (as many plies as )?
5. Composites
Warm-upHumanA company must make 5,000 carbon-fibre hydrogen pressure tanks a year: cylinders with domed ends, wrapped around a polymer liner. Which process fits best?
6. Composites
CoreHumanWhich of these are real, currently used or developing routes for handling retired glass-fibre/epoxy wind-turbine blades? Select all that apply.
7. Composites
CoreComputationalWrite off_axis_modulus(theta_deg, E1, E2, G12, nu12), which returns the Young's modulus of a unidirectional ply loaded at theta_deg degrees to its fibres, using
8. Composites
StretchComputationalIn vacuum infusion, Darcy's law gives the time for the resin front to travel a distance from a line inlet as . Write max_flow_length(gel_time, K, mu, phi, dP) that returns the longest distance (in metres) resin can travel before it gels, so that a process engineer can space the inlets. All inputs are in SI units (s, m², Pa s, dimensionless, Pa). Raise ValueError if any input is not positive.
Further reading and resources
- The elastic behaviour of a crystalline aggregate · R. Hill · 1952Proves that the Voigt and Reuss averages bound the true elastic moduli.
- A variational approach to the theory of the elastic behaviour of multiphase materials · Z. Hashin and S. Shtrikman · 1963Tighter bounds for isotropic mixtures.
- The Halpin–Tsai equations: a review · J. C. Halpin and J. L. Kardos · 1976
- The phenomena of rupture and flow in solids · A. A. Griffith · 1921Includes the experiments showing that thin glass fibres are far stronger than thick ones.
- Wind turbine blade waste in 2050 · P. Liu and C. Y. Barlow · 2017
- CMH-17 (Composite Materials Handbook)The industry handbook of composite material data, test methods and design practice for aircraft.
- WindEuropeThe European wind industry association, with reports on blade recycling and decommissioning.
- Composites UKA trade association with accessible guides to composite materials, processes and recycling.
- Mechanics of Composite Materials · R. M. JonesA standard textbook on micromechanics and classical laminate theory.