Rule of Mixtures — Longitudinal Modulus

Take a bundle of stiff fibres, surround them with a soft resin, pull along the fibres, and ask what stiffness comes back. The answer is the most quoted relation in composites and one of the most misused: $E_1 = E_f V_f + E_m (1 - V_f)$. Each phase contributes its own modulus in proportion to the volume it occupies.

The derivation is one line and the assumption inside it is the whole story. Fibre and matrix are bonded, so when the ply stretches they must stretch together — the same strain in both. That is the iso-strain assumption, and Voigt made it in 1889 for a quite different purpose. If the strain is shared, the stress in each phase is its own modulus times that strain, and the total force is the area-weighted sum. Divide back out by the total area and the volume fractions appear.

What makes it an upper bound rather than a prediction is everything the assumption quietly asserts. The fibres are perfectly straight. They are perfectly aligned with the load. The bond is perfect and no load is lost getting into them. Real laminae come within a few percent of that in longitudinal tension, which is why the rule survives at all — it is genuinely accurate in exactly one direction of one loading sense. In longitudinal compression it is optimistic, sometimes badly: real fibres are slightly wavy, and under compression a wavy fibre kinks sideways into the resin long before it reaches its own strength, so a carbon ply is typically good for only 50 to 70 percent of its tensile capability in the same direction.

Notice what the equation says about who is carrying the load. With a fibre fifty or eighty times stiffer than the resin, a 60% laminate has the fibres carrying something like 99% of the stiffness while occupying 60% of the volume. The resin is very nearly a passenger. That is why the simple average works here and, crucially, why it works nowhere else — turn the ply ninety degrees and the resin becomes the only path the load has, and the arithmetic inverts into the inverse rule of mixtures.

The classic mistake is entering a weight fraction where a volume fraction belongs. Burn-off and acid-digestion tests weigh the fibre residue and report $W_f$; every equation on this page is written in $V_f$. Carbon fibre is about one and a half times as dense as cured epoxy, so a 70% laminate by weight is only about 61% by volume — nine percentage points apart. Type 0.70 where 0.609 belongs and the longitudinal modulus comes out roughly 15% high, in the unsafe direction, and nothing in the answer looks suspicious because both numbers are dimensionless fractions in a plausible range. The conversion has a page of its own, because it needs both densities and cannot be done in your head.

The second classic mistake is using this number for a load that is not along the fibres. It is worth being blunt: a unidirectional carbon/epoxy ply is around nineteen times stiffer along its fibres than across them. That is not a correction factor, it is the material. Any formula with a single $E$ in it — axial deformation, beam deflection, Euler buckling — describes an isotropic solid, and pointing it at a lamina answers a question the material was never asked. Real structures are laid up in several directions precisely so that the laminate as a whole has usable properties in more than one, and the price is that a quasi-isotropic lay-up gives back more than half the longitudinal stiffness the fibres could have provided.

Run the equation backwards and it becomes useful in a different way. Solving for $E_f$ is how fibre moduli are actually obtained, since nobody tensile-tests a seven-micron filament if a coupon will do — and what comes back is an effective modulus that quietly absorbs misalignment, waviness and voids, which is more useful for predicting the next panel from the same process than the true filament figure would be. Solving for $E_m$, by contrast, is the weakest of the four directions: the resin contributes so little longitudinally that its modulus is a small difference between two large numbers, and a one-percent error in $E_f$ can move the answer by twenty. Get the matrix modulus from a transverse coupon instead, where the resin dominates and the arithmetic runs in your favour.