Eigenvalue sensitivity Created 2026-09-24 Updated 2026-11-05
For unit left eigenvector and right eigenvector of a real matrix at a simple eigenvalue , its normwise sensitivity isIt measures the first-order amplification of a matrix perturbation into an eigenvalue perturbation.
Nonorthogonal vectors Created 2026-09-24 Updated 2026-11-05
Two vectors are nonorthogonal when their inner product is nonzero. For the power method applied to a Hermitian matrix, this condition detects whether the starting vector has a nonzero component along a chosen eigenvector, because the eigenvectors can be taken as an orthonormal basis.
For a general nonnormal matrix, the relevant component is instead detected by a corresponding left eigenvector. For example,The dominant right eigenvector has nonzero dot product with , but , so repeated multiplication never picks up the eigenvalue two. The corresponding left eigenvector is , and correctly detects the missing dominant component. Thus an arbitrary right-eigenvector nonorthogonality condition is insufficient for nonnormal matrices.
In vector form the system is , withFor , its two real distinct eigenvalues are , with left eigenvectors . Thus the system is strictly hyperbolic there; at the eigenvalues coincide and the matrix is defective, so strict hyperbolicity fails. The characteristic directions are .
If , multiplication of by the left eigenvector gives . Since , defineThenConsequently is constant on the minus characteristic curves, and is constant on the plus characteristic curves. One reconstructs , with the essential physical constraint .
For the data on , the nonnegative square root givesThus, on any regular patch, use the two footpoints determined byand put , , selecting branches continuous from the initial point. These implicit algebraic formulas determine the local answer wherever both characteristic curve projections are regular and .
For example, near an initial point with , they give , , so . On an initial interval , define . One instead has andnear the chosen initial point . The radicands must be positive, the footpoints must stay in their indicated initial intervals, and must hold.
There is no continuously differentiable solution realizing the first data set on the entire stated initial line. At , , , and . The minus left eigenvector is and its eigenvalue is zero. The differential equation there would require , but its two sides are respectively and . This is an explicit compatibility contradiction, not merely a loss of a preferred formula. Also, the proposed simple expression independent of does not solve the equation for , because then but .
For the second data set, treat as the evolution variable and let label the initial coordinate. Then and . In the positive quadrant the minus characteristic curves give . The plus characteristic curves satisfy and . Therefore their continuous characteristic curve continuation isAbove the parabola the footpoint is ; below it the plus footpoint is . Each branch directly satisfies both equations and . They agree continuously on , but their derivatives are and , respectively. Thus the continuation is piecewise classical and not continuously differentiable across the parabola. The source's globally continuously differentiable formulation needs this qualification. If only positive initial is intended, the upper rational formula is a classical continuation throughout the positive quadrant, but the region requires incoming information from and is not determined by that positive half of the initial line alone. This illustrates loss of differentiability from square-root hyperbolic source data.
Past exam of the mathematics course of the University of Cambridge 2004 iii Paper 75 3 b Solution Created 2026-10-03 Updated 2026-10-14
Put and , the reduced gravity. Use the Boussinesq approximation, hydrostatic pressure, uniform vertical velocity and concentration profiles, negligible fluid entrainment and drag, and a flat bed. Integrating volume conservation, momentum conservation and the particle budget over the layer givesThe hydrostatic pressure resultant is ; the last right-hand side is the deposition flux. With , these areThe quasilinear coefficient matrix in variables isIts eigenvalues are , where . Therefore the characteristic curves and their compatibility ordinary differential equations areFor the last equation the left eigenvectors are , and multiplying the quasilinear equations by them gives the displayed source terms. This is the sedimenting rectangular-channel characteristic compatibility relation; it supplies one differential relation on each wave family, rather than three independently closed scalar equations there. The contact eigenvector has and . For reference, an equivalent form isOnly for constant without settling do become genuine Riemann invariants. With settling, the spatial buoyancy gradient cannot be silently omitted.
Past exam of the mathematics course of the University of Cambridge 2005 iii Paper 66 6 ii Solution Created 2026-10-03 Updated 2026-10-14
The corresponding subdivision surface interrogation needs mesh connectivity and patch adjacency, boundary and crease flags, local coordinate charts with transition maps, and subdivision into smaller patches with their original-chart mappings. Its evaluation enquiry returns a limit point, tangent derivatives and a normal vector; second derivatives are supplied only where the limit is sufficiently smooth. Certified bounding volumes and patch approximation errors allow a hierarchy for intersection and proximity searches. Boundary-curve enquiries complete the interface for clipping.
On regular parts of a subdivision surface, these evaluations may use its equivalent tensor-product or box spline patch. Near an extraordinary subdivision vertex, the local subdivision matrix determines the limit position through its constant left eigenvector, and its tangent modes determine the normal vector under the scheme's characteristic map of a subdivision surface regularity conditions. Refinement alone does not establish those conditions. A subdivision surface may have a well-defined tangent plane there without possessing second derivatives or Gaussian curvature. An interrogation interface must distinguish these cases and can work on surrounding regular charts when a higher derivative is unavailable at the vertex itself.
Together, subdivision, bounds, topology and limit evaluation support ray–subdivision surface intersection: bounds eliminate irrelevant patches; surviving patches are refined, and regular candidate intersections are corrected using position and first derivatives. The returned intersection is a point of the limit subdivision surface with an error certificate, rather than automatically a point on the last control mesh.
Past exam of the mathematics course of the University of Cambridge 2005 iii Paper 66 6 i Solution Created 2026-10-03 Updated 2026-10-14
A useful subdivision curve interrogation supplies the following connected capabilities. It describes the parameter interval, open or closed topology, endpoints and break points; it evaluates the limit point and, where the scheme guarantees them, and ; and it refines or restricts an interval while retaining the mapping to the original parameter. It also provides enclosing bounds and a certified curve-to-chord error on each restricted piece. These are enough to build clipping, intersections, closest-point searches, tangents, curvature and adaptive sampling.
The enquiries concern the subdivision limit, not just a finite refined control polygon. For a stationary local refinement matrix , a limit vertex can be obtained from a normalized left eigenvector , , as . Derivatives use the appropriately scaled difference schemes or equivalent spline evaluation on regular intervals. The convergence and derivative hypotheses must actually hold for the scheme; no second derivative can be requested at an intrinsic crease.
If the subdivision mask is nonnegative and preserves constants, descendants and limit points lie in the relevant control-point convex hull, giving a useful enclosing volume. For subdivision masks with negative weights, that hull need not be an enclosure: the interface must supply another valid bound, for example from a proven refinement-error estimate. This distinction is important for safe spatial rejection in an algorithm.
Past exam of the mathematics course of the University of Cambridge 2023 ii Paper 4 40C b Solution Created 2026-09-24 Updated 2026-09-25
Differentiate the eigenvalue equationat , where . Suppressing the argument givesLeft-multiplication by the transpose of the unit left eigenvector and use of cancel the terms involving :Because the eigenvalue is simple, , and hence the first-order perturbation of a simple eigenvalue isThe definition of the operator norm, followed by the Cauchy-Schwarz inequality, givesTherefore
Past exam of the mathematics course of the University of Cambridge 2023 ii Paper 4 40C c Solution Created 2026-09-24 Updated 2026-09-25
If is a real symmetric matrix, the finite-dimensional spectral theorem lets us choose the normalized left eigenvector and right eigenvector to be the same vector. Thus
More generally, the unitary diagonalization of a normal matrix shows that a real normal matrix also has coincident normalized left and right eigenspaces for each real eigenvalue. Since the question assumes , one can again choose , and every such simple eigenvalue has
Past exam of the mathematics course of the University of Cambridge 2023 ii Paper 4 40C d Solution Created 2026-09-24 Updated 2026-09-25
The displayed matrix is an upper bidiagonal matrix. For , the right eigenvector equation successively givesWe may therefore take the unit right eigenvector to be .
For a left eigenvector, givesSince ,Choosing the initial scale producesAfter normalization,The eigenvalue sensitivity is thereforewhere is the factorial. Hence
A left eigenvector satisfies . Here is a symmetric matrix, so its left and right eigenspaces have the same components. The characteristic polynomial factors asThe eigenvalues and all corresponding nonzero row eigenvectors areFor example, in the block, equal components give the eigenvalue and opposite components give , while the coordinate is unaffected.
Suppose is a diagonalizable matrix, , with an eigenbasis and spectral decompositionAssume the dominant eigenvalue is a simple eigenvalue andThen the normalized power iteration approaches the one-dimensional eigenspace of . Its asymptotic direction-error bound has factor . This spectral gap is a gap in modulus: if , then . A small relative gap gives slow convergence.[ref]
The short proof is the exact eigenbasis identityThe remaining sum has norm bounded by a constant times . This proves convergence of direction, allowing a changing sign or complex factor. A tiny or a poorly conditioned, nonorthogonal eigenbasis can delay visible convergence: the condition number of affects constants, while controls the eventual rate. The bound need not be an equality; if the initial coefficient of the second-largest eigenvalue vanishes, the active spectrum of the power method gives a faster rate. When , all unwanted components disappear after one multiplication.[ref]
The hypothesis is , not an ordinary nonzero inner product with the right eigenvector. Let be the first row of , a left eigenvector normalized so is for and otherwise; denotes conjugate transpose. Then . For a Hermitian matrix, choose an orthonormal eigenbasis and the criterion becomes : the starting vector must be nonorthogonal to the dominant eigenvector. For a general diagonalizable matrix, the nonorthogonal eigenbasis makes these two tests different, as illustrated under power iteration failure cases.
Power iteration failure cases 2026-10-14
A tied largest modulus can prevent even direction convergence. With and ,The two directions form a cycle; neither is an eigenvector. There is no dominant-modulus spectral gap. A tied modulus is not a universal failure: an initial eigenvector remains on its line, and preserves every direction. Subspace iteration addresses a separated cluster by targeting a dominant invariant subspace instead of one eigenvector.
An orthogonal start can fail despite a strict spectral gap. For and , the dominant component is zero, so and every iterate equals . The Rayleigh quotient is exactly and the eigenpair residual is zero, although the dominant eigenvalue is . Exact arithmetic never creates a missing component.[ref]
The third panel contrasts these failures with , , for whichHere the horizontal eigenvector direction converges with a strict spectral gap, while the normalized vectors alternate signs. This is the successful direction convergence described under sign and phase in power iteration.
For a nonorthogonal eigenbasis, even the word “orthogonal” must be used carefully. The diagonalizable matrix has right eigenvectors , , and dominant left eigenvector . For , the dominant coefficient is . The starting vector is nonorthogonal to but stays at the smaller eigenvalue ; is orthogonal to but has and converges to the dominant direction. The left eigenvector test, not the right-eigenvector inner product, controls the outcome.
A random vector drawn from a continuous probability distribution with a probability density function on the full real or complex vector space has with probability one. This avoids the exact missing-component hyperplane, but does not guarantee a large component or fast convergence. Floating-point arithmetic may seed a missing component, yet a structurally preserved zero can remain zero, so accidental perturbations are not a convergence guarantee.
Power method Created 2026-09-24 Updated 2026-10-14
Iterate . For a self-adjoint operator with a unique dominant eigenvalue separated in modulus from the others, a starting vector that is nonorthogonal to its dominant eigenvector gives convergence of the eigendirection at the spectral-ratio rate. For a general diagonalizable matrix, the condition is a nonzero coefficient in the dominant component of its eigenbasis, detected by the corresponding left eigenvector; an ordinary dot product with the right eigenvector does not generally detect that coefficient.
The power method is power iteration; its convergence depends on the dominant spectral component of the starting vector.
For example, has dominant right eigenvector and a dominant left eigenvector . The starting vector has nonzero dot product with the right eigenvector, but : it stays forever in the eigenspace for the smaller eigenvalue . Its coefficient along the dominant direction is zero, as the left eigenvector detects. An orthonormal eigenbasis makes the two tests agree in the self-adjoint case.
Right eigenvector Created 2026-09-24 Updated 2026-11-05
A right eigenvector of a square matrix is a nonzero column vector satisfying for some scalar . The scalar is its eigenvalue. Requiring is essential, since the zero vector satisfies the displayed equation for every .
For example,Thus the two displayed vectors are right eigenvectors with different eigenvalues. A left eigenvector instead acts on from the left; specifying the side matters for a general matrix.
Stationary distribution Created 2026-09-24 Updated 2026-10-14
A stationary distribution is a probability distribution preserved by a Markov chain's evolution. For a Discrete-time Markov chain in the row convention,Thus is a normalized nonnegative left eigenvector of the transition matrix for eigenvalue one. Starting with distributed as makes every have that same probability distribution. For a time-homogeneous Markov chain, the path probabilities are also unchanged by shifting all observation times together: the process has strict stationarity. This does not mean its individual paths are constant. See Hao Wu's MIT lecture 2, stationary distributions.
For the sunny/rainy transition matrix, write . The sunny component of gives , so andMoreover,These equal stationary flows verify detailed balance and show that the Markov chain is reversible. The diagonal detailed balance equations hold automatically. Only of the stationary days are rainy, but of the days following a rainy day are rainy: stationarity permits dependence.
A finite Markov chain that is an irreducible Markov chain has a unique stationary distribution. If it is also an aperiodic Markov chain, for every initial probability distribution. The sunny/rainy example satisfies both conditions: both directions are possible and each state has a positive holding probability of a Markov chain. See Hao Wu's MIT lecture 4, convergence theorem. For an infinite state space, the additional positive recurrence hypothesis in the countable-state Markov chain convergence theorem matters.
Existence of a stationary distribution alone does not give this convergence. Consider the deterministic two-cycleHere , but starting in the first state gives alternating probability distributions . Its period of a state in a Markov chain is two. The running fraction of visits to either state nevertheless tends to , illustrating the Markov chain ergodic theorem. Starting with instead keeps the one-time probability distributions stationary even though every path alternates.
Strict hyperbolicity 2026-10-31
A first-order system is strictly hyperbolic at a state when has distinct real eigenvalues. These separate its characteristic families and permit independent compatibility equations from left eigenvectors. At a repeated eigenvalue strict hyperbolicity fails; if the matrix is also defective, even hyperbolicity in the complete-eigenvector sense fails. For the eigenvalues are distinct precisely when .

