Why the stress-strain curve appears to drop after the ultimate tensile strength — even though the material never stops getting stronger.
Every stress-strain curve you've ever plotted from a tensile test relies on one quiet assumption: that the cross-sectional area carrying the load is the area you measured before the test started. That assumption is engineering stress — load divided by the original area, A₀, a fixed number that never changes for the rest of the calculation. It's simple, it's standard, and it's what every published stress-strain curve and design allowable is built on. But it isn't what's physically happening inside the specimen. True stress divides the same load by the actual, instantaneouscross-sectional area — the area the material actually has at that exact moment, which shrinks continuously in tension. Early in a test the two numbers are nearly identical. Near necking, they diverge sharply — and the direction of that divergence is exactly what produces one of the most persistently misread features of the tensile test: the apparent post-peak "weakening" that isn't weakening at all.
Engineering stress, σ = F/A₀, divides the instantaneous load by the specimen's originalcross-sectional area, measured once before loading begins and never updated. That's not an approximation error — it's a deliberate simplification, and it's the right one for almost every practical purpose. A₀ is easy to measure, easy to design around, and every allowable stress, safety factor, and published stress-strain curve you'll encounter in a materials handbook is built on it. The trade-off is that once the specimen's actual area stops matching A₀ — which happens continuously, but only becomes significant later in the test — engineering stress stops describing the true intensity of loading the material is experiencing.
True stress, σ_true = F/A, divides the same instantaneous load by the specimen's actual cross-sectional area at that moment — the area after whatever reduction has already occurred. In tension, a ductile specimen is continuously thinning as it stretches, so A keeps shrinking and true stress keeps climbing relative to engineering stress for a given load. Through the elastic region and most of the plastic region, the area change is small enough that σ and σ_true stay close together. Once necking begins — deformation localizing into one narrow region instead of staying spread uniformly along the gauge length — the area at that neck starts shrinking rapidly, and the gap between the two curves opens fast.
Once necking starts, the reduction in cross-sectional area at the neck outpaces whatever additional load the still-hardening material can support. Divide that load by the original, no-longer-accurate area A₀, and the calculated engineering stress goes down — not because the material is carrying less force capability, but because the fixed denominator can no longer keep up with what's really happening at the neck. Divide the same load by the actual, shrunken area A at the neck, and true stress keeps climbing all the way to fracture, because it's tracking the real, ever-increasing intensity of load on an ever-shrinking cross-section.
Up until necking begins, deformation stays uniform along the whole gauge length, and if you assume the material's volume stays constant (a good assumption for metals well into the plastic region), the two stresses relate directly: σ_true = σ(1 + ε), where ε is the engineering strain. That's why the curves track closely — true stress is always somewhat higher, by a small and predictable amount, but the relationship is clean and calculable. The instant necking begins, that uniform-deformation assumption stops holding: strain localizes into the neck instead of staying spread along the gauge length, so the simple formula no longer applies anywhere near the neck. From that point on, true stress has to be found from the actual measured area at the neck, not calculated from the engineering curve — and because the neck's area is now shrinking far faster than the rest of the specimen, the two curves pull apart sharply rather than gradually.
No — the material is doing the opposite. It keeps work-hardening all the way to fracture; that's exactly what the true stress curve, which never stops rising, is showing you. The apparent drop in engineering stress is purely a bookkeeping artifact: past necking, the load is still being divided by A₀, the original area — a number that stopped describing the specimen's real cross-section the moment the neck started forming. The load-carrying capacity of the whole specimen does eventually fall, because the neck's area is collapsing faster than the material's hardening can compensate for — but that's a statement about the shrinking cross-section, not about the material's intrinsic resistance to further deformation. Read the true stress curve, and there is no point where the material "gets weaker" — it is at its strongest, per unit of actual area, right at the moment of fracture.
Explains the difference between engineering stress (load divided by the original, fixed cross-sectional area, A₀ — the basis for standard stress-strain curves and design allowables) and true stress (load divided by the actual, instantaneous cross-sectional area, which shrinks continuously in tension) — and why the two track closely until necking begins, then diverge sharply, producing the characteristic post-UTS peak-then-decline shape of the engineering curve even as the material keeps work-hardening.
The engineering stress-strain curve peaks at the ultimate tensile strength and then visibly declines on its way to fracture, and it's natural to read that decline as the material weakening. It isn't. The decline is a direct consequence of dividing an evolving load by a denominator, A₀, that was only ever accurate before necking began. Once the neck's actual area starts shrinking faster than the load, that fixed denominator can no longer represent what's physically happening at the neck, and the calculated engineering stress falls — even while the material itself continues to work-harden and requires an ever-increasing true stress to keep deforming it further.
Before necking, deformation is uniform along the gauge length, and assuming the material's volume stays constant, true stress and engineering stress relate directly by σ_true = σ(1 + ε), where ε is engineering strain — a small, predictable, calculable gap. Necking marks the point where deformation stops being uniform and localizes into one narrow region; from that point on, the uniform-strain assumption behind that formula no longer holds anywhere near the neck, so true stress must be obtained from the actual measured area at the neck rather than calculated from the engineering curve. Because the neck's area then collapses rapidly while the material is still hardening, true stress keeps climbing all the way to fracture, while engineering stress — still using the stale original area — appears to fall.
Design allowables, published material-property tables, and the everyday stress-strain curve virtually always use engineering stress, because A₀ is easy to measure and gives simple, repeatable numbers — appropriate for structures operating well within the elastic region, far from necking. True stress and true strain matter wherever a process pushes material through large plastic deformation with significant area change: metal-forming simulations (forging, extrusion, deep drawing), finite element models of ductile failure, and any flow-stress curve used to predict material behavior deep into the plastic region. Feeding an engineering stress-strain curve into a large-deformation forming simulation without converting to true stress and true strain first is a common source of simulation error, since the engineering curve's post-necking shape doesn't reflect the material's real flow behavior.
Because engineering stress is calculated by dividing the load by the specimen's original, fixed cross-sectional area (A₀). Once necking begins, the specimen's actual area at the neck shrinks faster than the load can keep rising, so the load itself starts to fall — and dividing that falling load by the unchanged A₀ produces a falling engineering stress value. It's a feature of the calculation's fixed denominator, not evidence that the material has become intrinsically weaker.
No — for a ductile metal undergoing necking, true stress rises continuously all the way to fracture, because the material keeps work-hardening (requiring more stress to continue deforming it) even as its cross-sectional area at the neck keeps shrinking. There is no point on the true stress-strain curve where the material's resistance to further deformation, per unit of actual area, goes down.
Yes, but only before necking begins. Assuming the material's volume stays constant during uniform plastic deformation, true stress and engineering stress are related by σ_true = σ(1 + ε), and true strain relates to engineering strain by ε_true = ln(1 + ε). Once necking starts, deformation is no longer uniform along the gauge length, so these formulas stop being valid — true stress after that point has to come from directly measuring the actual neck area, not from converting the engineering curve.
Engineering stress, for essentially all conventional structural design. Structures are designed to stay well within the elastic region, far below necking, where engineering and true stress are close enough that the distinction rarely matters, and every standard material property table, allowable stress, and safety factor is defined in engineering-stress terms. True stress becomes important in specialized contexts — metal-forming process design and finite-element simulations of large plastic deformation — where the material is deliberately pushed far into the plastic region and the original-area assumption breaks down.
Necking begins at the ultimate tensile strength, the point where the rate of strain hardening (the material getting stronger as it deforms) can no longer keep pace with the rate at which the cross-sectional area is shrinking. Up to that point, any local, slightly-thinner region hardens enough to resist further thinning faster than the rest of the specimen, keeping deformation uniform. Past that point, a locally thinner region can no longer out-harden its own area loss, so deformation concentrates there instead of staying spread out — that localized, accelerating thinning is the neck.
Try our Materials Science & Metallurgy Studio
More calculators, simulators, and guides for this discipline.