What is a p-y curve?

A p-y curve is the nonlinear relationship between soil resistance p — force per unit length of pile — and lateral pile deflection y at one specific depth. The p-y method models a laterally loaded pile as a beam supported by a stack of these independent nonlinear springs, one per depth, and it is the basis of COM624P, LPILE, RSPile, and PileCalc.

A beam on nonlinear springs

Picture the pile as a vertical beam and the soil as a series of horizontal springs attached along its length — the classic Winkler idealization, except the springs are not linear. Push the pile sideways a little and the soil responds stiffly; push harder and the response softens until the soil reaches its ultimate resistance and flows around the shaft. Each spring's force-deflection law is a p-y curve, and it changes with depth: near the surface a passive wedge of soil can heave upward, so resistance is low, while at depth the soil is confined and must flow around the pile, so resistance is much higher.

The pile itself obeys beam-column mechanics. Combining the two gives the governing fourth-order differential equation of the p-y method:

EI·y⁗ + Pₓ·y″ + Eₚᵧ·y − W = 0

where EI is the pile's flexural rigidity, Pₓ is the axial load (which amplifies deflection — the P-delta effect), W is any distributed lateral load, and Eₚᵧ = p/y is the secant modulus of the p-y curve at the current deflection. Because Eₚᵧ depends on y, the equation is nonlinear and has no closed-form solution for real soil profiles — it has to be solved numerically, which is why the p-y method lives in software rather than in hand calculations. The full formulation is documented in Wang & Reese (1993), FHWA-SA-91-048, the COM624P manual, and in our lateral pile analysis docs.

Where p-y curves come from

The standard curve families are not theoretical constructs — they were back-calculated from full-scale instrumented lateral load tests. Matlock (1970, OTC 1204) derived the soft clay criteria from tests on 324 mm steel pipe piles at Lake Austin and the Sabine River. Reese, Cox & Koop (1974, OTC 2080) derived the sand criteria from tests on 610 mm pipe piles at Mustang Island, Texas, and the same team published the submerged stiff clay criteria a year later (1975, OTC 2312).

The procedure was elegant: the test piles carried strain gauges down their length, so the bending moment profile M(x) was measured directly. Differentiating M twice gives the soil reaction p; integrating M/EI twice gives the deflection y. Cross-plotting p against y at each gauge depth produced experimental p-y curves, and the published criteria are analytical expressions fitted to those measurements, written in terms of routine soil parameters so engineers can construct the curves at any site.

The main p-y curve families

Each family pairs a soil type with the load test that calibrated it. These are the models PileCalc implements, and the parameters each one needs:

As a concrete example, the Matlock soft clay model caps the resistance at the lesser of a shallow wedge and a deep flow-around mechanism,

pᵤ = min[ (3 + γ′·x/c + J·x/b)·c·b , 9·c·b ]

and builds the static backbone as a cube-root curve, p = 0.5·pᵤ·(y/y₅₀)^(1/3), where y₅₀ = 2.5·ε₅₀·b is the deflection at half the ultimate resistance. Every family follows the same pattern — an ultimate resistance from a failure mechanism, and an empirical backbone shape fitted to the load tests.

How a solver actually uses the curves

The p-y curves themselves are only half the method; the other half is the numerical solution. The pile is discretized into finite-difference nodes (PileCalc defaults to 100 increments), and the solver iterates:

• Assume a deflected shape y(x) — a small seed deflection on the first pass.
• At each node, evaluate the local p-y curve at the current y and form the secant modulus Eₚᵧ = p/y.
• Solve the now-linear beam equation with those spring stiffnesses for a new deflected shape.
• Repeat — with under-relaxation to stabilize the iteration — until the deflections stop changing (PileCalc's default tolerance is a relative change of 10⁻⁶).

This is Picard iteration on the secant modulus, exactly the scheme COM624P used. It converges in tens of iterations for routine problems. Failure to converge is itself diagnostic: it usually means the soil has reached its ultimate resistance over most of the pile and the system is approaching geotechnical failure.

Static versus cyclic curves

Every classic family comes in two flavors. Static curves represent monotonic, first-time loading. Cyclic curves represent the degraded response after repeated load reversals — wave loading, wind, traffic — and they matter most in clays, where cycling remolds the soil and can open a gap around the pile near the surface. Matlock's cyclic criteria cap the resistance at 0.72·pᵤ at depth and degrade it further above a critical depth; Welch & Reese shift the stiff clay curve as a function of the number of cycles; the API sand model reduces its plateau factor A to 0.9. If the design load reverses, analyze with cyclic curves — the difference in deflection can be substantial at high load levels.

What p-y curves don't capture

Honest limits, because the method has them:

• Empirical basis. The curves were calibrated on a small number of full-scale tests, mostly on piles under about a meter in diameter. Extrapolation to very large monopiles is an active research topic, and offshore wind practice has moved toward site-specific calibration.
• Independent springs. Winkler springs don't talk to each other, so the method ignores continuum shear transfer between depths. In layered profiles this shows up at layer boundaries; the Georgiadis (1983) equivalent-depth construction — which LPILE, RSPile, and PileCalc all offer — corrects the ultimate resistance carried across layers.
• Group shadowing. A pile in a trailing row pushes into soil already disturbed by the row ahead, so its p-y curves must be scaled down by p-multipliers. How those factors work is covered in our pile group efficiency post.
• Dimensional inputs. Parameters like the subgrade modulus k (force/length³) are dimensional, and published correlation tables are tied to specific unit systems — a classic source of order-of-magnitude errors. See units and conventions before mixing sources.

How this differs from the Broms method

Broms (1964a/b) is the other name engineers meet first. It is an ultimate-strength hand method: idealize the soil as purely cohesive or purely frictional, assume a rigid or plastic-hinge failure mechanism, and read the ultimate lateral capacity from charts. It answers "what load fails this pile?" — quickly, and without a computer. The p-y method answers the harder, more useful question: "what are the deflection, moment, and shear at my working load?" — with real layering, nonlinearity, axial load, and cyclic degradation. Broms remains a fine independent check on a p-y analysis; the full comparison is in Broms method vs p-y method.

Common questions

Is a p-y curve the same as a spring constant?

No. A spring constant is a single linear stiffness, but a p-y curve is a full nonlinear force-deflection relationship: stiff at small deflections, softening toward an ultimate resistance. Solvers linearize it locally as a secant modulus Eₚᵧ = p/y, but that modulus changes at every depth and every load level — which is why lateral pile response is nonlinear even for an elastic pile.

What does y₅₀ mean in a p-y curve?

y₅₀ is the deflection at which the soil mobilizes half its ultimate resistance, and it sets the horizontal scale of clay p-y curves. It comes from ε₅₀, the laboratory strain at 50% of the maximum principal stress difference: Matlock takes y₅₀ = 2.5·ε₅₀·b, so a softer clay (larger ε₅₀) or a bigger pile diameter b stretches the curve toward larger deflections.

Do I need lab tests to build p-y curves?

Ideally yes — c and ε₅₀ from triaxial tests, φ′ from lab or in-situ testing. In practice, preliminary curves are often built from SPT correlations (for example, undrained strength and ε₅₀ estimated from blow count N). Correlations are approximate by nature, so treat those runs as sensitivity studies and firm up the parameters that the answer actually depends on.

Are p-y curves valid for large-diameter monopiles?

With caution. The classic criteria were calibrated on piles roughly 0.3–0.6 m in diameter, and rigid, low-slenderness monopiles mobilize resistance components — base shear, side friction moments — the standard curves omit. Recent offshore wind practice supplements or replaces API curves with site-calibrated ones. For conventional pile diameters, the classic families remain the validated standard.

The fastest way to build intuition is to look at actual curves: our free p-y curve generator plots any of the families above from your own soil parameters, at any depth, static or cyclic — no signup required.