Eccentric Column (Secant Formula)

stabilitystressmass-cost

The Euler column is a beautiful idealization: perfectly straight, loaded exactly through the centroid, perfectly content until — at one critical load — it bifurcates. Real columns never get that story. A bracket bolts on a few millimeters off-axis; a crane column carries its rail on one flange; even “centered” loads ride manufacturing tolerances. Offset the load by an eccentricity $e$ and the column carries a bending moment $M = P(e + v)$ from the first newton — it bows immediately, the bow lengthens the moment arm, the longer arm deepens the bow. The self-amplification has an exact elastic answer, the secant formula:

$\sigma_{max} = \frac{P}{A}\left[1 + \frac{ec}{r^2}\,\sec\!\left(\frac{L}{2r}\sqrt{\frac{P}{EA}}\right)\right]$

Two things to read out of it. First, the dimensionless eccentricity ratio $ec/r^2$ is the only place the offset enters — it families every column design chart ever printed. Second, the secant’s argument reaches $\pi/2$ exactly when $P = P_E = \pi^2 EI/L^2$: the deflection and the stress diverge as the load approaches the Euler load. The imperfect column never buckles in the clean eigenvalue sense — it just runs away asymptotically toward the limit its perfect twin would have reached. (Dial $P$ upward in the widget and watch the bow do exactly this.)

Now the point this page exists to make. The stress is nonlinear in the load — double $P$ and $\sigma_{max}$ more than doubles, because the secant grows too. So the familiar margin $\sigma_y/\sigma_{max}$ — fine for every linear THING in this catalog — overstates your safety. The honest question is: what load first yields a fiber? That means solving

$\sigma_y = \frac{P_y}{A}\left[1 + \frac{ec}{r^2}\,\sec\!\left(\frac{L}{2r}\sqrt{\frac{P_y}{EA}}\right)\right]$

for $P_y$ — and no algebra will ever isolate it: $P_y$ sits both outside and inside the secant. Shigley’s text says it plainly: the load “appears twice … this equation needs an iterative process.” This page does exactly that, live — the root is bracketed inside $(0, P_E)$, where it provably exists and is unique (the residual falls monotonically from $\sigma_y$ to $-\infty$), and a verified root-finder pins it on every knob drag. The build proves the bracket’s sign change at dozens of sampled states and cross-checks the browser’s roots against 60-digit bisection. Both margins are displayed side by side: $\mathrm{SF}_\sigma$ (deceptive) and $\mathrm{SF}_P$ (honest). At the default state they disagree by about 17% — in the unsafe direction.

One more cascade worth clicking through the material picker: swap A36 for quenched 4340 at the same $E$ and the stress readouts do not move — elastic stress amplification is set by stiffness and geometry alone — but $P_y$ jumps by multiples. Strength is back in the story precisely because failure here is yielding, not the geometric escape of the perfect column.

Try it

Governing relations

Derivation

Steps marked modeling step are where physics enters by citation — every other line is machine-proven to follow from them, and the cited models are independently re-derived in the test pipeline where possible. See what is and isn't machine-verified.

How it fails

The secant formula is exact elastic theory for an idealized member — and each idealization marks a way the real column quits earlier:

• First yield is the beginning of the end, not a nuisance. Past $P_y$ the yielded fiber stops taking extra stress, the effective section shrinks, the bow deepens, and the moment arm grows — the same amplification loop that built the secant formula now feeds on a softening column. Collapse follows at a load not far above first yield, which is why $P_y$ (and not some plastic reserve) is the design ceiling this page reports. The widget refuses the elastic readouts past first yield rather than pretend the fiction continues; $P_y$ and $\mathrm{SF}_P$ stay exact.
• The margin must be taken on load. A stress-based factor of safety silently assumes proportionality. Here a “factor of 2 on stress” can evaporate into far less than a factor of 2 on load — the page shows both margins so the gap is impossible to miss. Codes write this rule down (apply the factor to the load, then check the stress); Shigley states it as the reason the secant equation must be solved iteratively at all.
• Approaching $P_E$ is hopeless even though nothing has “buckled.” The deflection diverges asymptotically; stiffness against the bow goes to zero. Real eccentric columns are serviceable only well below the Euler load — by the time $P$ reaches even $0.7\,P_E$, the amplification factor is brutal and growing without bound. There is no event to wait for, just runaway.
• Both-ends-same-side bowing is the worst case this page models. Equal eccentricities on the same side give single-curvature bending with the peak at midspan. Opposite-side offsets (reverse curvature) are gentler; a single-end offset lands between. The pinned, equal-offset case is the standard conservative chart.
• Tension can appear on the convex face. Push the eccentricity ratio up and the bending gradient overwhelms the uniform compression — the neutral axis walks into the section and part of it goes into tension (watch the zero-stress line enter the bar in the sim). Bolted bases, grouted joints and masonry — things that carry compression but not tension — fail this way long before any fiber yields.
• Crooked columns are eccentric columns in disguise. An initial bow $a$ behaves nearly like an end eccentricity; design codes fold real-world crookedness, residual stress and accidental offset into equivalent-imperfection rules precisely because the perfectly straight, perfectly centered member of the Euler page does not exist in any shop.

Sources

• Gere, J. M., & Goodno, B. J., Mechanics of Materials, 9th ed., Cengage, 2018 — §11.5 (Columns with Eccentric Axial Loads) and §11.6 (The Secant Formula for Columns).
• Budynas, R. G., & Nisbett, J. K., Shigley's Mechanical Engineering Design, 10th ed., McGraw-Hill, 2015 — §4-13 (Columns with Eccentric Loading): the secant equation for the critical load, and the design rule that the factor of safety must be applied to the load because the stress is a nonlinear function of it.