Principal axes and principal values

The formula

The defining eigenproblem, written for stress:

σ · n̂ = σp · n̂

Reading: on a plane with normal n̂, the traction vector points straight along n̂ — that plane carries pure tension or compression and zero shear. The directions n̂ that satisfy this are the principal axes; the multipliers σp are the principal values.

Expanding the eigenproblem gives the characteristic equation:

σp³ − I₁·σp² + I₂·σp − I₃ = 0
I₁ = trace(σ)      I₂ = ½·(trace(σ)² − trace(σ²))      I₃ = det(σ)

Reading: a cubic whose three roots are the principal stresses. The coefficients I₁, I₂, I₃ are the invariants — they come out the same in any coordinate frame, which is why failure criteria are built from them.

The 2-D closed form every mechanical engineer memorizes:

σ₁,₂ = (σx + σy)/2 ± √( ((σx − σy)/2)² + τxy² )
tan(2θp) = 2·τxy / (σx − σy)

Reading: the average normal stress is the center of Mohr's circle, the square root is its radius, and θp is how far to rotate the element before the shear terms vanish.

Maximum shear, which lives at 45° to the principal axes:

τmax = (σ₁ − σ₃)/2

Reading: half the spread between the largest and smallest principal stresses — all three, not just the two in your plane.

The guarantee underneath all of it (the spectral theorem): a real symmetric matrix always has real eigenvalues and mutually perpendicular eigenvectors. Symmetry is what makes the whole scheme work.

Where you meet it

At the strain-gauge rosette on a test article. Three gauges at 0°/45°/90° give εx, εy, γxy; the principal-strain formulas turn that into magnitudes and a direction. When the principal direction comes out 20° off the loading axis, the load path is not what the model said — that number has started more than one hardware investigation.
At the stress review board. FEA post-processors report von Mises for ductile margins and max principal for brittle materials, welds, and composite plies. Both are functions of the principal values, and the reviewer asking "max principal or von Mises?" is asking which failure physics you believe.
In the mass-properties lab. A spin-balanced missile or satellite must spin about a principal axis of its inertia tensor. Nonzero products of inertia mean the geometric axis and the principal axis disagree, and the residual tilt shows up as coning and bearing load on the spin table. Balancing is the act of driving the products of inertia to zero.
At the bench with a shaft under bending plus torsion. Neither the bending stress nor the shear stress alone predicts where the crack starts; the principal stress from the combined state does — which is why fatigue cracks in shafts spiral at roughly 45°.

How it works

Start with why symmetry matters. Moment equilibrium on a differential element forces τxy = τyx, so the stress tensor is symmetric, and the spectral theorem then promises three real principal values with perpendicular directions. No complex roots, no skewed axes, ever. If the matrix you are diagonalizing is not symmetric, you are not solving this problem — check your bookkeeping before your solver.

A worked case, numbers verified: σx = 80, σy = 20, τxy = 40 (MPa). Center (80+20)/2 = 50, radius √(30² + 40²) = 50. So σ₁ = 100 MPa, σ₂ = 0 MPa, at θp = ½·atan(80/60) = 26.57°. Rotate the element 26.57° and the shear stress is gone — 100 MPa of pure tension on one face, nothing on the other. Same material, same load, same point; only the description changed. That is the entire content of "principal": the state was never partly shear, it was pure tension viewed from a tilted frame.

The invariants are the free sanity check. I₁ = σx + σy + σz must equal σ₁ + σ₂ + σ₃ to machine precision in any frame. When a hand calc and an FEA output disagree, compare traces first — it localizes the error in thirty seconds.

The mistakes, in the order they actually cost people:

Forgetting the third principal stress. In plane stress the out-of-plane principal value is zero, and it still counts. If both in-plane principals are tensile — say 100 and 40 MPa — the true τmax is (100 − 0)/2 = 50 MPa on a plane tilted out of the sheet, not the in-plane (100 − 40)/2 = 30 MPa. Pressure-vessel skins fail exactly there, and the 2-D formula quietly hides it.
The tan(2θp) ambiguity. The arctangent returns two roots 90° apart, and the formula does not tell you which one carries σ₁. Substitute back, or read it off Mohr's circle. Labeling the minor axis as the major one puts your crack-growth prediction perpendicular to the actual crack.
Engineering versus tensor shear strain. Rosette data gives γxy; the tensor wants γxy/2. Skip the factor of two and the principal strain directions come out wrong while the magnitudes look plausible — the worst kind of error.
Repeated principal values. In a hydrostatic or axisymmetric stress state, two (or three) eigenvalues coincide and any direction in the shared plane is principal. Solvers then return arbitrary vectors, and a principal-direction vector plot over that region looks like static. That is the math being honest, not the code being broken.
Assuming stress and strain share axes. They do in isotropic material. In a composite laminate the stiffness tensor rotates the response, and the principal stress and principal strain directions part ways. Instrumenting a composite as if it were aluminum bakes this error into the data.

The inertia version is the same mathematics with different letters: diagonalize the inertia tensor and the products of inertia vanish along the principal axes. The dynamics adds one twist the stress problem lacks — spin about the axis of intermediate moment of inertia is unstable. Flip a phone or a hammer and watch it: the tennis-racket theorem, live. But be careful taking the theorem straight to flight hardware. Rigid-body theory permits stable spin about either the major or the minor axis; a real spacecraft is not rigid. Fuel slosh and flexible appendages dissipate energy, and at fixed angular momentum the only state of minimum kinetic energy is spin about the major axis — so minor-axis spin is only marginally stable and, given any dissipation, decays into a flat tumble. Explorer 1 taught the lesson in 1958: the ABMA/JPL satellite went up spinning about its long, minor axis, its flexible whip antennas bled off energy, and within hours it was tumbling end over end about its major axis ^[11]^[12]. Long-duration spinners have followed the major-axis rule ever since; minor-axis spin appears only briefly, on spinning upper stages held straight by nutation control until burnout.

History

The problem arrived through spinning tops before anyone wrote a matrix. Johann Andreas Segner, working at Göttingen, showed that every rigid body possesses three mutually perpendicular free axes of rotation, publishing the result in his Specimen theoriae turbinum of 1755 ^[1]^[2]. Segner knew Euler and corresponded with him; Euler then built the axes into the foundation of rigid-body dynamics, and his 1765 Theoria motus corporum solidorum seu rigidorum derives the equations of motion in exactly the frame where the inertia tensor is diagonal ^[2]^[3]. Lagrange later recognized what these directions were in modern terms: the eigenvectors of the inertia matrix ^[4].

Stress joined the story in the 1820s. Augustin-Louis Cauchy presented his memoir on internal forces in a continuum to the Paris Academy in 1822, establishing that the traction on any internal plane comes from a single symmetric array of nine components — the stress tensor — with the details published over the following years, including De la pression ou tension dans un corps solide in 1827 ^[5]^[6]. The planes of pure normal traction fell out of his construction: principal stresses were born as a by-product of defining stress correctly.

Cauchy then proved the theorem that makes engineers' lives easy. In his 1829 memoir on the secular inequalities of planetary motion he showed that the roots of what he named the équation caractéristique of a symmetric array are always real — the origin of both the result and the phrase "characteristic equation" ^[7]^[8]. And in 1882 Christian Otto Mohr, following a graphical-statics suggestion Karl Culmann had made in 1866, published the circle diagram that turned the 2-D eigenproblem into a compass-and-straightedge exercise every strength-of-materials student still draws ^[9]^[10].

Related tools

Sources

Written by HE in our own words from the cited sources — engineering judgment included, your stamp still required. All entries →