Find your next learning step
When two tensor eigenvalues nearly coincide, the reported axes can rotate sharply while the physical state changes very little. Recovery needs measures that respect that uncertainty.
Open the chapter contents · Polymer Science root · Gold Standard of Education · How X Works
Chapter contents
Choose the route closest to the question, or read from Chapter 1 for the complete argument. The chapters remain open on the page; these grouped routes are an index.
Route 1 · Define unstable axes
Route 2 · Work the tensor ledger
Route 3 · Test gaps and objectivity
Route 4 · Check limits and evidence
Route 5 · Ask and continue
CHAPTER 1 OF 12 · DEFINE UNSTABLE AXES
1. Notice when a principal axis stops being a stable object
A symmetric positive-definite conformation tensor can be described by eigenvalues and orthogonal eigenvectors. The eigenvalues measure extension along principal directions. The eigenvectors supply the directions. That description is useful only while the relevant eigenvalues are sufficiently separated from one another.
When two eigenvalues are equal, any orthonormal basis of their shared eigenspace is valid. When they are merely close, a small measurement or numerical perturbation can rotate the reported axes sharply even though the tensor changes very little.
CHAPTER 2 OF 12 · DEFINE UNSTABLE AXES
2. Separate eigenvalue gaps from physical recovery
Let λ1 ≥ λ2 ≥ λ3. The gaps λ1 − λ2 and λ2 − λ3 tell us how well individual axes are separated. A small gap makes the corresponding eigenvectors sensitive. Davis–Kahan perturbation theory formalises the role of the gap: subspace error can be bounded by perturbation size divided by spectral separation.
The stable object near degeneracy is often the combined eigenspace, not either named vector inside it. Reporting a dramatic axis flip without the gap can turn coordinate ambiguity into a false story of molecular reorientation.
CHAPTER 3 OF 12 · WORK THE TENSOR LEDGER
3. Work a nearly degenerate tensor ledger
Consider C_A = diag(4, 2.01, 2.00). Its trace is 8.01 and determinant is 16.08. The first axis is clearly separated; the second and third are divided by only 0.01. Now rotate the last two coordinate directions by 45 degrees while keeping the same eigenvalues. The resulting C_B has the same trace, determinant and eigenvalue set.
The named second and third axes have moved, but the two-dimensional subspace they span has not. A reader who tracks individual arrows sees a 45-degree event. A frame-invariant reader sees no change in the spectrum and no change in total stretch.
| Quantity | C_A | C_B | Interpretation |
|---|---|---|---|
| trace | 8.01 | 8.01 | same total squared extension measure |
| determinant | 16.08 | 16.08 | same volume-like tensor invariant |
| λ2 − λ3 | 0.01 | 0.01 | both have weak in-plane axis separation |
| reported in-plane axes | coordinate aligned | rotated 45° | basis choice changes |
CHAPTER 4 OF 12 · WORK THE TENSOR LEDGER
4. Build frame-invariant recovery measures
Useful scalar summaries include trace, determinant or log-determinant, eigenvalue ratios, and the Frobenius norm of the deviatoric part C − tr(C)I/3. Distances defined on the geometry of positive-definite tensors can compare configurations without choosing a laboratory frame.
No single invariant tells the entire recovery story. Trace can return while anisotropy remains. Determinant can agree while eigenvalue distributions differ. Report a small, declared set whose physical meaning matches the experiment.
| Measure | Insensitive to rigid frame rotation? | Question answered |
|---|---|---|
| trace | yes | How much total extension remains? |
| determinant or log-determinant | yes | How has the product of principal stretches changed? |
| deviatoric norm | yes | How far from isotropic is the tensor? |
| individual eigenvector angle | no near degeneracy | Where is a well-separated axis pointing? |
CHAPTER 5 OF 12 · WORK THE TENSOR LEDGER
5. Use projectors when axes are nearly tied
For the near-degenerate pair, form the projector P23 onto the subspace spanned by eigenvectors two and three. Compare projectors between times rather than matching the two vectors individually. The projector is unchanged by swapping axes or rotating the basis within the subspace.
If the leading eigenvalue remains separated, its one-dimensional projector can still be tracked. This mixed description—one stable leading direction plus one stable transverse plane—is more faithful than forcing three stable arrows.
CHAPTER 6 OF 12 · TEST GAPS AND OBJECTIVITY
6. Changed context: the spectral gap opens
Replace the eigenvalues (4, 2.01, 2.00) with (4, 2.8, 1.2) while keeping a perturbation of the same size. The gaps are now large. A 45-degree jump in the second axis would require a much larger tensor change and should no longer be dismissed as a nearly degenerate basis choice.
This transfer check shows why an angle needs its eigenvalue gap. The same observed angle can mean weak identifiability in one state and strong reorientation in another.
CHAPTER 7 OF 12 · TEST GAPS AND OBJECTIVITY
7. Connect axis uncertainty to restart overshoot
After cessation, a second deformation can be imposed parallel, transverse or oblique to the remembered structure. If the recovered tensor has a well-separated leading axis, restart geometry can be defined relative to it. If all eigenvalues nearly coincide, fine directional labels are weakly supported.
Continue through Tensorial Entanglement Recovery, Eigenframes and Direction-Dependent Overshoot and Orthogonal Double-Step Strain, Tensor Memory and Cross Components for restart protocols.
CHAPTER 8 OF 12 · TEST GAPS AND OBJECTIVITY
8. Distinguish objectivity from eigenvector stability
Objectivity asks whether a constitutive description transforms correctly under a change of observer. Eigenvector stability asks whether small perturbations change an inferred direction. A frame-invariant scalar can be objective yet insensitive to important directional information; an objective tensor can still have nearly degenerate, unstable eigenvectors.
Use both checks. First ensure the quantity is not an artefact of coordinate choice. Then quantify whether the data resolve the spectral structure being interpreted.
CHAPTER 9 OF 12 · CHECK LIMITS AND EVIDENCE
9. Limits and common misconceptions
An eigenvalue tie does not mean the material has no structure; it means directions inside that eigenspace are not distinguished by this tensor. Stable invariants do not prove microscopic recovery, and an apparently stable eigenvector can be biased by regularisation or temporal smoothing.
The conformation tensor is a reduced description. Different molecular distributions can share it. Measurement noise, closure assumptions, spatial averaging and loss of positive definiteness can all affect the inferred spectrum.
CHAPTER 10 OF 12 · CHECK LIMITS AND EVIDENCE
10. Build an evidence check
Archive the tensor before eigendecomposition, units, symmetrisation rule, positive-definiteness check, spatial and temporal averaging, eigenvalues, gaps, projectors, invariant measures and uncertainty method. Preserve the raw signal needed to repeat the calculation.
Perturb tensors within estimated error and recompute axes. Report angle distributions only for separated directions; report projector distances for clustered eigenvalues. Compare conclusions using trace, log-determinant and deviatoric norm, then test whether restart stress predictions improve.
CHAPTER 11 OF 12 · ASK AND CONTINUE
11. Questions readers should ask
Does a 90-degree eigenvector flip prove a 90-degree material rotation?
No. Near a repeated eigenvalue, vector labels can swap or rotate with very small tensor change.
Are eigenvalues themselves frame-invariant?
Yes under orthogonal changes of laboratory coordinates, although their estimates still carry measurement and model uncertainty.
Why not report only the trace?
Trace loses directional and anisotropy information. It should be paired with other justified invariants or stable subspaces.
CHAPTER 12 OF 12 · ASK AND CONTINUE
12. Continue through verified routes
Begin with the Polymer Science root, Gold Standard of Education and How X Works. Continue through Shear Cessation, Elastic Recoil and Structural Recovery, Waiting-Time Spectra, Restart Superposition and Directional Recovery and Interrupted Shear, Stress Memory and Rejuvenation Cycles.
Primary routes include Hameduddin and colleagues’ geometric decomposition of the conformation tensor, Davis and Kahan’s eigenvector perturbation analysis, and Advani and Tucker’s tensor description of orientation. The nearly degenerate tensor ledger is an original teaching example.
Top · Contents · Previous · How X Works