The ideal slender column
Pinned at both ends, straight, uniform, elastic. This is an equation instrument, not a claimed laboratory measurement.
Euler calculation ready.
Pcr = π²EI / (KL)², with K = 1
Physical seam · a solved equation meets a real wall
The Rocket Wall That Distrusts Its Own Theory
A perfect thin cylinder has a beautiful classical buckling load. Real shells can fail far below it. Turn the radius-to-thickness dial, compare a slender column with a shell, and test the imperfection-sensitivity curve that helps explain why a conservative NASA lower bound from 1968 still appears in the 2020 design guidance.
The surprise is not that the mathematics is unsolved. The surprise is that rolling a flat elastic wall into a cylinder makes its perfect solution dangerously sensitive to the tiny shape errors that every manufactured wall has.
Layer one · operate the gapSame elasticity, two very different promises
A slender, straight, elastic, pin-ended column has a clean critical load. A perfect isotropic cylinder under uniform axial compression also has a clean critical stress. The second answer is not a safe prediction for an imperfect shell, so NASA multiplies it by an empirical lower-bound factor, γ.
The thin cylindrical shell
Unstiffened, isotropic, unpressurised, elastic, and under uniform axial compression. Those limits are load-bearing.
Shell calculation ready.
σcl = 0.605 E(t/R) · σdesign = γσcl
The plotted five-specimen set is from Wang et al. (2019), at approximately the same R/t = 333. It shows a spread across five tests with differing imperfection treatments: W1-W2 used measured imperfections, while W3-W5 also used deliberately introduced actuator or dimple imperfections. This is not ordinary repeat-test scatter or the older archive from which the 1968 lower bound was fitted.
Layer two · put in the flawA dent smaller than the wall is thick
The instrument below reproduces the Koiter small-deviation relation printed in the 2020 revision of SP-8007. The slider is the imperfection amplitude divided by wall thickness. It is a particular normalized imperfection model at Poisson ratio ν = 0.3, not a universal prediction for every dent.
Koiter imperfection sensitivity
The perfect shell begins at one. Move even slightly away from perfect geometry and the normalized load falls along the sourced curve.
Imperfection calculation ready.
γ = 1 + αε − √[αε(2 + αε)] · α = ¾√[3(1 − ν²)]
The flaw, deliberately magnified
The drawn waviness is exaggerated so it remains visible. The number below is the actual slider amplitude.
Actual model input: ε = 0.10. Drawing waviness exaggerated.
At zero imperfection the relation returns the classical load. At one tenth of a wall thickness, it does not return nine tenths. The square-root structure makes the first movement away from perfection steep. That is the point of imperfection sensitivity: “small geometry error” and “small strength error” are not synonyms.
This curve explains a mechanism. It does not replace a detailed nonlinear analysis, measured imperfection map, boundary-condition model, material model, or test. The 2020 guidance discusses those newer paths and retains lower-bound recommendations as guidelines.
The check: every live result recomputed here
The browser and the repository verifier use the same equations independently. Change any slider and this line is rebuilt from the current inputs:
Computing the live check…
- Measured input: none of the three calculators measures a physical specimen. Inputs are user-selected geometry plus published or explicitly representative constants.
- Model values: column modulus and shell modulus are both set to 70 GPa, representative of aluminium. The column width is fixed at 30 mm. These are free choices, not measurements.
- Published equations: Euler's pinned-column formula, the classical cylinder stress, the SP-8007 empirical factor, and the Koiter small-deviation relation are evaluated live. The coefficient 0.605 assumes ν = 0.3.
- Published test data: the five cyan points are source values 0.886, 0.887, 0.633, 0.645, 0.569 at approximately R/t = 333. W1-W2 used measured imperfections; W3-W5 also used deliberately introduced actuator or dimple imperfections. The points are plotted as sourced, not generated from the SP curve.
- Range and limits: the old empirical relation was recommended for R/t < 1500 and with caution for L/R > 5. This page stays below the first limit and does not model the second variable.
- Open and configuration-dependent: the exact collapse load of a real shell depends on its imperfection field, load introduction, boundary conditions, material variation, plasticity, cutouts, stiffening, and pressure. The single R/t curve cannot know those things.
Repository check: node research/the-rockets-missing-two-thirds/verify-the-rockets-missing-two-thirds.mjs
What the curve is, and what it is not
Verified
The 1965 publication was revised in August 1968. NASA issued a second revision in 2020, and the current revision still prints the empirical axial-compression relation.
Modelled
The calculators assume elastic, isotropic, unstiffened, unpressurised geometry. The Koiter slider uses one normalized imperfection form and a chosen Poisson ratio.
Not claimed
The page does not say every rocket loses two thirds, that the worst historical ten-percent result was typical, or that one curve sizes every modern vehicle wall.
Why “as low as ten percent” is not the headline number
A 2023 study that compiles 514 experimental results notes that the SP-8007 guideline approaches a lower-bound factor of 0.1 at large shell slenderness. A 2020 methods paper describes most collected factors as roughly 0.4 to 1, with some below 0.2. The extreme is historically important, but treating it as the ordinary shell would be false. The live SP curve reaches 0.179 at R/t = 1500, not 0.1.
Why pressure and stiffeners are outside this instrument
Internal pressure can stabilize a cylinder, while stringers, rings, sandwich cores, composite layups, cutouts, joints, and load introduction change the relevant stiffness and buckling modes. SP-8007 contains other sections and other recommendations for several of those cases. Applying this unstiffened, isotropic, axial-compression curve to all of them would be a category error.
Did the 2020 revision “solve” shell buckling?
No. The second revision incorporates work from NASA's Shell Buckling Knockdown Factor assessment and discusses modern analysis and design approaches. It explicitly presents itself as a guideline, not a NASA requirement unless a program makes it one, and advises designers to follow updates in the state of the art. Better measurement and nonlinear analysis can produce less conservative, configuration-specific estimates. They do not abolish imperfection sensitivity.
Primary apparatus
- V. I. Weingarten, P. Seide, and J. P. Peterson, Buckling of Thin-Walled Circular Cylinders, NASA SP-8007, September 1965, revised August 1968. Equations 5 and 6 give the empirical axial-compression factor and state its range and length-ratio caution.
- Mark W. Hilburger, Buckling of Thin-Walled Circular Cylinders, NASA/SP-8007-2020/REV 2, second revision, December 2020. Equations 7 through 9 reproduce the classical stress and empirical factor. Equations 138 and 139 give the Koiter small-deviation relation used in layer two.
- B. Wang et al., Experimental validation of cylindrical shells under axial compression for improved knockdown factors, International Journal of Solids and Structures 164 (2019), 37–51. Table 7 supplies the five plotted test factors.
- R. M. J. Groh and A. Pirrera, Probing the stability landscape of cylindrical shells for buckling knockdown factors, Philosophical Transactions of the Royal Society A 381 (2023). The paper identifies 514 results from 20 studies spanning more than a century and discusses the lower-bound factor near 0.1.
- X. Ma, P. Hao, F. Wang, and B. Wang, Incomplete reduced stiffness method for imperfection sensitivity of cylindrical shells, Thin-Walled Structures 157 (2020), 107148. This methods paper reports the collected knockdown-factor range cited above.
Corrections
Corrected 2026-09-29: the SP-8007 citation named “J. P. Peterson, Jr.”; NASA lists “J. P. Peterson”. Found by the source-first pilot, 2026-09-28. Corrected 2026-09-29: Groh and Pirrera (2023) was called a review; it is a research article that compiles 514 experimental results. Found by the source-first pilot, 2026-09-28. Corrected 2026-09-29: the image alt text described a buckling cylinder, dimples and a load curve; the image is a title card, and its alt text now gives the layer's title. Found by the source-first pilot, 2026-09-28. Corrected 2026-09-29: the drawing caption claimed 18× magnification at ε = 0.10; that multiplier did not match the drawing's scale, so the caption now says the waviness is exaggerated. Found by the source-first pilot, 2026-09-28. See the assay.