Why slender steel columns fail before they yield - and what EC3 actually checks

Most early-career engineers size a column by comparing the axial load to the squash load, $N_{pl,Rd} = A f_y / \gamma_{M0}$. For stocky members that is fine. But the moment a column becomes slender, a different failure mode takes over long before the steel ever reaches yield: flexural buckling. A perfectly straight, perfectly elastic strut does not crush - it bows sideways at a load that depends on stiffness and length, not on strength.

The elastic critical load

The starting point is Euler's result for the load at which an ideal pin-ended strut becomes unstable:

$N_{cr} = \frac{\pi^2 E I}{L_{cr}^2}$

Everything important is in that equation. Resistance grows with the flexural stiffness $EI$ but falls with the square of the effective length $L_{cr}$. Double the length and you quarter the buckling load. Because $I$ is smallest about the minor axis, an unrestrained column almost always buckles about its weak axis - which is why the minor-axis radius of gyration $i_z$ usually governs.

The effective length $L_{cr}$ is the real length multiplied by a factor for the end conditions: roughly 1.0 for pinned-pinned, 0.7 for pinned-fixed, 2.0 for a cantilever. Getting this factor wrong is one of the most common ways a column check goes astray.

The non-dimensional slenderness

Real columns are neither perfectly straight nor perfectly elastic, so EC3 does not use $N_{cr}$ directly. Instead it forms the non-dimensional slenderness:

$\bar\lambda = \sqrt{\frac{A f_y}{N_{cr}}}$

This single number places the member on a buckling curve. A low $\bar\lambda$ means a stocky column that fails near its squash load; a high $\bar\lambda$ means a slender one governed almost entirely by stability.

The imperfection factor α is not a fudge - it encodes the residual stresses and out-of-straightness that every real column carries from rolling and fabrication.

EC3 provides several buckling curves (a0, a, b, c, d), each with its own imperfection factor α. Which curve applies depends on the section shape, the axis of buckling, the steel grade and how the section was made. The reduction factor χ is then read from the curve at the calculated $\bar\lambda$, and the design check becomes:

$N_{Ed} \le \chi \, N_{pl,Rd}$

The practical meaning of χ is a discount on the squash load: a stocky member keeps almost all of its resistance (χ near 1.0), while a slender one may keep only a fraction.

Why this governs more often than expected

The lesson is simple but easy to forget: for anything but the stockiest members, the buckling check is the design. Three habits keep you safe:

1. Use the right effective length. Bracing, base fixity and the direction of restraint all change $L_{cr}$, and the answer scales with its square.
2. Check the weak axis. Unless a column is restrained about its minor axis at close centres, that is the axis that fails first.
3. Do not chase strength to fix a buckling problem. A higher grade barely helps a slender column - the failure is governed by stiffness and geometry, not by $f_y$. Shortening the effective length or picking a stockier section is far more effective.

Putting numbers on it

The formulas above are the whole method, but the arithmetic - $N_{cr}$ from the minor-axis inertia, then $\bar\lambda$, then χ off the right curve - is where mistakes creep in. Pull the exact $A$, $I_z$ and $i_z$ for any size straight from the steel section tables, then open the section in the steel catalogue and run an EC3 capacity check to get the buckling resistance for your effective length without doing it by hand. The companion section modulus explainer covers the bending side of the same member.