Column Buckling

Why columns buckle

Press down on a metre-long ruler held vertical and notice that it bows sideways well before it crushes. The buckling load is governed not by material strength but by the column's geometric stiffness against lateral deflection. Once the axial load exceeds a critical threshold, any tiny imperfection — initial out-of-straightness, slight eccentricity of the load — gets amplified, and the column snaps sideways into a curved shape.

Euler derived the threshold by writing the differential equation for a slightly bowed column under axial load P:

EI · d²y/dx² + P · y = 0

Solving with pinned-pinned boundary conditions (y = 0 at x = 0 and x = L)
gives sinusoidal solutions y = A · sin(nπx/L), with the smallest non-trivial
load at n = 1:

   P_cr = π² · EI / L²    (pinned-pinned, mode 1)

For other end conditions, replace L with KL (effective length):

   P_cr = π² · EI / (KL)²

Buckling modes

The same equation has higher solutions where n = 2, 3, 4, ... — these are the mode shapes. Mode n has n half-waves along the column, n−1 internal inflection points, and a critical load n² times mode 1. For a pinned-pinned column of fixed length:

Mode n	Half-waves	Inflection points	P_cr	Shape
1	1	0	π²EI/L²	Single bow (most common)
2	2	1	4π²EI/L²	S-shape, requires brace at midpoint
3	3	2	9π²EI/L²	W-shape, two midpoint braces
4	4	3	16π²EI/L²	Four half-waves, three braces
5	5	4	25π²EI/L²	Five half-waves, four braces
n	n	n−1	n²·π²EI/L²	Higher modes need more braces

In practice, an unbraced column always buckles into mode 1 because the moment it reaches P_cr,1 it loses stability and never approaches the mode-2 load. Higher modes only become relevant when intermediate braces force them — a column braced at midspan can't use mode 1 and so it reaches the mode-2 load, four times higher. A column braced at thirds reaches mode 3, nine times higher. Bracing is the cheapest way to raise capacity.

End conditions and K

End conditions	K theoretical	K practical (AISC)	P_cr (relative)	Effective length	Example
Pinned-pinned	1.0	1.0	1.0×	L	Truss compression member
Fixed-fixed	0.5	0.65	4.0× (theoretical)	0.5L	Welded steel column in moment frame
Fixed-pinned	0.7	0.80	2.0× (theoretical)	0.7L	Concrete column with pinned cap
Fixed-free (cantilever)	2.0	2.10	0.25×	2L	Flagpole, billboard post
Pinned-free (translation)	1.0 (sway)	2.0+	varies	varies	Sway-permitted frame column
Sway frame, both fixed	1.0 (sway)	1.2	1×	L	Multi-storey building

The "practical" K values are inflated above the theoretical ones because real "fixed" connections aren't perfectly rigid — bolted base plates rotate slightly, weld groups stretch under moment. AISC and Eurocode practice values bake in this realism. Using theoretical K can overestimate capacity by 20–30%.

Worked example: Euler critical loads at K = 1, 0.7, 2

Consider a steel column 3 m long, hollow circular section with outer diameter 100 mm and wall thickness 5 mm. Compute critical load for three end conditions:

Section properties:
  Outer radius r₀ = 0.050 m
  Inner radius r_i = 0.045 m
  I = π(r₀⁴ − r_i⁴)/4
    = π(6.25×10⁻⁶ − 4.10×10⁻⁶)/4
    = π × 2.15×10⁻⁶ / 4
    = 1.69×10⁻⁶ m⁴

E = 200 GPa = 2 × 10¹¹ Pa
EI = 2 × 10¹¹ × 1.69×10⁻⁶ = 338,000 N·m²
L = 3 m

Case 1 — Pinned-pinned (K = 1):
  KL = 3 m
  P_cr = π² × 338,000 / 9 ≈ 370,700 N = 371 kN

Case 2 — Fixed-pinned (K = 0.7):
  KL = 2.1 m
  P_cr = π² × 338,000 / 4.41 ≈ 756,500 N = 756 kN

Case 3 — Fixed-free / cantilever (K = 2):
  KL = 6 m
  P_cr = π² × 338,000 / 36 ≈ 92,700 N = 93 kN

The same physical column has a critical load that ranges over a factor of 8 (93 kN to 756 kN) just by changing how the ends are connected. This is why end-condition assumptions are the most consequential decision in column design.

Slenderness ratio and the Euler-Johnson transition

The Euler formula gives an unrealistically high critical stress for short, stocky columns — it would predict P_cr / A > σ_yield, which is impossible because the material yields first. The boundary between elastic Euler buckling and inelastic material yielding is the critical slenderness:

λ = KL / r          where r = √(I/A) is the radius of gyration

λ_c = π · √(E/σ_y)  is the critical slenderness ratio

For mild steel (E = 200 GPa, σ_y = 250 MPa):
  λ_c = π · √(800) = π · 28.3 ≈ 89

Two regimes:
  λ > λ_c  — elastic Euler buckling: σ_cr = π²E / λ²
  λ < λ_c  — Johnson short-column parabola:
             σ_cr = σ_y − (σ_y · λ / (2π√(E/σ_y)))²

The Johnson formula tangentially joins the Euler hyperbola at λ = λ_c, providing a smooth design curve from very stocky (pure yield) through intermediate (mixed) to slender (pure Euler). Most modern codes (AISC, Eurocode 3) use slightly different empirical curves with the same shape.

Beyond pure Euler

• Eccentric loading (secant formula). Real loads are never perfectly centred. Adding eccentricity e gives an exact closed-form: σ_max = (P/A)(1 + (ec/r²) · sec((KL/2r)·√(P/EA))). Even tiny e (1–5 mm) can drop capacity 20–40%.
• Imperfect columns (Perry-Robertson). Adding an initial sinusoidal bow and solving gives the basis for European steel design code formulas — a smooth curve below pure Euler that captures real-world capacity.
• Inelastic / tangent-modulus theory. For columns at the elastic-plastic boundary, replace E with the tangent modulus E_t. Engesser's tangent-modulus theory and Shanley's reduced-modulus theory bracket the actual behaviour.
• Local buckling. Thin-walled sections can buckle locally — flange or web wrinkling — before global Euler buckling. Code limits on b/t ratios prevent this.
• Plate / shell buckling. Thin plates and shells (rocket interstages, pressure vessels) have more complex buckling behaviour governed by the Donnell equations.

Common failure modes

• Eccentric loading triggers premature buckling. The dominant cause of in-service buckling failures: a load applied off-centre by even a few millimetres creates a starting moment that amplifies under load. Always check the secant formula in conditions where load eccentricity is plausible.
• Out-of-straightness reduces capacity. Manufacturing tolerances allow steel columns to be initially bowed by L/1000 — Perry-Robertson knockdown puts real capacity at perhaps 70–85% of pure Euler.
• Weak-axis buckling. A wide-flange column has different I about its two axes; if you forget which axis is unbraced, you'll size for strong-axis I and the column will buckle weak-axis.
• Local flange or web buckling before global. Thin walls in cold-formed shapes and built-up plate columns can wrinkle locally first; the global column then loses effective area and buckles at a much lower load.
• Connection rotation softens K. A "fixed" base plate that actually rotates slightly under moment shifts effective K from 0.5 toward 0.7 or 1.0; design codes catch this with practical K values.
• Imperfect bracing. Intermediate braces require minimum stiffness and minimum strength to actually force a higher mode shape; flimsy bracing doesn't raise capacity at all.