Structural
Castigliano's Second Theorem: Deflection as the Derivative of Strain Energy
Point a screw at a cantilever, hang a 5 kN load off its tip, and you can predict the beam's deflection to the nearest tenth of a millimeter without ever drawing a free-body diagram of internal forces — you just differentiate the stored elastic energy with respect to that load. That is the power of Castigliano's Second Theorem, published by Italian railway engineer Carlo Alberto Castigliano in his 1873 thesis at the Polytechnic of Turin.
The theorem states that in a linearly elastic structure, the displacement at the point and in the direction of an applied load equals the partial derivative of the total strain energy with respect to that load: δ = ∂U/∂P. Its rotational twin gives slope from a moment: θ = ∂U/∂M. It converts a geometry problem into a calculus problem, and it remains a workhorse for hand-analysis of beams, trusses, frames, and curved members.
- TypeEnergy method for elastic deflection
- Key equationδ = ∂U/∂P ; θ = ∂U/∂M
- InventedCarlo A. Castigliano, 1873 (Turin thesis)
- Used inBeams, trusses, frames, curved bars, springs
- Strict requirementLinear-elastic material, small deflections
- Bending strain energyU = ∫ M²/(2EI) dx
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
What It Is and Where It's Used
Castigliano's Second Theorem is an energy method for finding the displacement of a linearly elastic structure at a specific point. Rather than integrating a differential equation of the elastic curve, you write the total strain energy U stored in the structure as a function of the applied loads, then take a partial derivative with respect to the load of interest to recover the deflection in that load's direction.
- Beams and frames — tip deflection of cantilevers, mid-span sag of portal frames.
- Trusses — joint displacement from axial member forces, U = Σ N²L/(2AE).
- Curved members — crane hooks, chain links, and C-clamps, where bending dominates.
- Statically indeterminate structures — treat a redundant reaction R as an unknown load; since the support does not move, ∂U/∂R = 0 gives the compatibility equation.
It is taught in every mechanics-of-materials course and still used for quick, auditable hand-checks of FEA output. Castigliano's First theorem is the inverse: force equals the derivative of energy with respect to displacement.
How It Works — The Derivation
Consider an elastic body loaded by forces P₁, P₂, … Pₙ. The strain energy U equals the external work done as loads rise slowly from zero. Because the material is linear-elastic, U depends only on the final load values, not the path. Now increase one load Pᵢ by dPᵢ. The extra work stored is (∂U/∂Pᵢ) dPᵢ.
Reverse the loading order: apply dPᵢ first, then the full set P. The dPᵢ does external work dPᵢ·δᵢ as point i moves through its final displacement δᵢ. Equating the two paths (Clapeyron's theorem guarantees they match) gives:
- δᵢ = ∂U/∂Pᵢ — displacement at i, along Pᵢ.
- θᵢ = ∂U/∂Mᵢ — rotation at i, if Mᵢ is a couple.
For bending, U = ∫ M²/(2EI) dx, so δ = ∫ (M/EI)(∂M/∂P) dx. The elegance is that you differentiate before integrating: ∂M/∂P is simply the coefficient of P in the internal moment expression, which keeps the algebra light.
Key Quantities and a Worked Example
Take a cantilever of length L = 2 m carrying a point load P = 5 kN at its free tip. Section: steel I-beam, E = 200 GPa, I = 8.0 × 10⁻⁶ m⁴, so EI = 1.6 × 10⁶ N·m².
- Measuring x from the free end, the internal moment is M = −P·x, so ∂M/∂P = −x.
- δ = ∫₀ᴸ (M/EI)(∂M/∂P) dx = ∫₀ᴸ (−Px)(−x)/EI dx = (P/EI) ∫₀ᴸ x² dx = P·L³ / (3EI).
Plugging in: δ = (5000 × 2³) / (3 × 1.6 × 10⁶) = 40 000 / 4.8 × 10⁶ = 8.33 × 10⁻³ m = 8.3 mm. This exactly reproduces the textbook PL³/3EI result — Castigliano derives it without assuming the formula.
If you needed deflection at a point with no real load, you add a fictitious dummy load Q, carry it symbolically through ∂M/∂Q, then set Q = 0 after differentiating.
Using It in Practice
The practical recipe is short and mechanical, which is why it is favored for exams and design audits:
- 1. Write internal actions — express M(x), N (axial), and V (shear) in terms of the real loads plus any dummy load Q needed at the target point.
- 2. Differentiate — compute ∂M/∂P (or ∂M/∂Q), which are just the load coefficients.
- 3. Integrate — evaluate δ = Σ ∫ (M/EI)(∂M/∂P) dx over each segment; add axial term ∫ N(∂N/∂P)/AE dx and, for stubby members, the shear term.
- 4. Set dummy loads to zero and read off the displacement.
For trusses the integral collapses to a sum: δ = Σ (Nᵢ Lᵢ / AᵢE)(∂Nᵢ/∂P). Shear strain energy is usually below 3% of the total for slender beams (L/h > 10) and is routinely dropped; keep it for deep beams. Sign convention matters: a positive δ means the point moves in the same direction as the assumed load.
Comparison to Related Methods
Castigliano's Second Theorem is mathematically equivalent to the unit-load (virtual work) method — differentiating M²/(2EI) with respect to P produces exactly the m·M/EI integrand of the unit-load approach, where m is the moment from a unit dummy force. Engineers often view Castigliano as the energy-formalism sibling of virtual work.
- vs. double integration — Castigliano gives one number at one point cheaply; double integration gives the whole curve but bogs down with multiple loads and discontinuities.
- vs. moment-area / conjugate beam — those are graphical shortcuts tied to the M/EI diagram; Castigliano is purely analytic and extends naturally to curved and 3D members.
- vs. FEM — finite elements handle arbitrary geometry but hide the physics; Castigliano remains the transparent hand-check that catches a mis-entered modulus or moment of inertia.
Crucially, all these classical methods share Castigliano's linear-elastic, small-deflection assumptions.
Limits, Failure Modes, and Significance
The theorem's assumptions are also its failure modes. It is valid only when superposition holds, which means:
- Linear-elastic material — no yielding; steel stresses must stay below ~250 MPa for mild steel. Plastic hinges void the derivation.
- Small deflections — geometry is evaluated in the undeformed configuration; large-displacement (P-delta) effects break it.
- No initial stresses or nonlinear supports — thermal pre-stress, gaps, and follower forces require extensions.
A subtle trap is applying it to displacements where no load acts and forgetting the dummy load, or differentiating after substituting numeric loads (you must differentiate symbolically first). Despite these limits, its significance endures: it unifies beam, truss, and frame deflection under one calculus operation, underpins the force (flexibility) method for indeterminate structures via ∂U/∂R = 0, and gave engineers a rigorous energy foundation decades before matrix and finite-element methods existed.
| Method | Best for | Handles indeterminacy? | Key limitation |
|---|---|---|---|
| Castigliano's 2nd theorem | Point deflection/slope in beams, trusses, curved members | Yes (via redundant reaction as unknown) | Needs a load at the point (dummy load if none) |
| Unit / virtual load method | Same class; often identical integrals | Yes | Requires two force systems (real + virtual) |
| Double integration | Full deflection curve of simple beams | Poorly | Tedious with multiple loads/discontinuities |
| Moment-area method | Slopes and deflections from M/EI diagram | Limited | Graphical, error-prone for complex loading |
| Conjugate beam | Deflections via analogy loading | Limited | Requires constructing an analog beam |
| Finite element method | Any geometry, large/complex structures | Yes | Software-dependent; overkill by hand |
Frequently asked questions
What is the difference between Castigliano's first and second theorems?
The second theorem gives displacement as the partial derivative of strain energy with respect to a force: δ = ∂U/∂P. The first theorem is its inverse — force equals the derivative of energy with respect to a displacement: P = ∂U/∂δ. The second theorem is far more commonly used because loads are usually the known quantities and deflections the unknowns.
Why does Castigliano's theorem require a linearly elastic material?
The derivation equates strain energy with external work and relies on load superposition, which only holds when stress is proportional to strain. If the material yields or behaves nonlinearly, the energy stored depends on loading path and the simple derivative relationship fails. In that case Crotti–Engesser's theorem, using complementary energy, must be used instead.
What is the dummy load method and when do I need it?
When you want the deflection at a point where no external load is applied, Castigliano's derivative ∂U/∂P has nothing to differentiate against. You add a fictitious 'dummy' load Q at that point in the direction of interest, carry it symbolically through the moment expression, differentiate, then set Q = 0. The result is the true deflection produced by the real loads alone.
Do I need to include shear and axial strain energy?
It depends on geometry. For slender beams and frames (span-to-depth ratio greater than about 10), bending strain energy dominates and shear contributes under 3%, so it is usually neglected. For deep beams, short brackets, and trusses, axial energy N²L/2AE is essential, and shear energy matters for stubby members. Always include the terms that carry significant force in your specific structure.
How is Castigliano's theorem used for statically indeterminate structures?
Treat each redundant reaction R as an unknown external load and write the strain energy in terms of it. Because a rigid support does not move, its displacement is zero, so ∂U/∂R = 0. This yields one compatibility equation per redundant, which you solve for the unknown reactions — the basis of the force (flexibility) method.
Is Castigliano's theorem the same as the unit load method?
They are mathematically equivalent for linear-elastic structures and produce identical integrals. Differentiating the bending energy M²/2EI with respect to load P gives the integrand (M/EI)(∂M/∂P), where ∂M/∂P plays the exact role of the unit-load moment m in virtual work. Castigliano frames it as an energy derivative; virtual work frames it as two interacting force systems, but the answer is the same.