For sine collocation of square modified Helmholtz global relations, each opposite-side mode has a scaled matrix
The assembled system has strict diagonal dominance, and its spectral condition number of a positive-definite matrix is below . The conclusion concerns the explicit paired sine rows, not arbitrary uncombined complex rows.
Consider the original linear system
Completing the square proves that is a symmetric positive-definite matrix:
The Jacobi preconditioner has . Symmetric positive-definite preconditioning gives
Its eigenvectors and have eigenvalues and . Hence
The characteristic polynomial of is . Its eigenvalues and spectral condition number of a positive-definite matrix are
The ellipse axis ratio drops from about to . The transformation multiplies the second coordinate by ten, bringing the two energy scales together.
Figure 1.
Diagonal scaling changes thin energy ellipses into nearly circular energy ellipses
. Details.
Both panels show the same energy level sets, expressed in error coordinates and . Their coordinate axes have equal scales. The marked displacement becomes ; its energy remains . These are exact contours of the specified quadratic form.
Solving the original equations gives
Direct matrix multiplication checks both linear-system residuals:
By contrast, left preconditioning gives , which is visibly not a symmetric matrix, despite having the same positive eigenvalues as .
A bad scaling can reverse the benefit: the still valid symmetric positive-definite matrix gives . Its Rayleigh quotients on the coordinate unit vectors are and , so . Cheapness and positive definiteness do not guarantee improved conditioning.
For a real positive-definite matrix in finite dimension, write by the spectral theorem for real symmetric matrices. In the coordinates in an orthonormal basis , the positive level set becomes
This is an ellipsoid, with semi-major and semi-minor axes in two dimensions forming an ellipse; each semi-axis has length and direction given by the corresponding eigenvector. In the opening example, the shorter semi-axis is along , while the longer is along . A stiffer direction, meaning a larger eigenvalue, reaches the same energy at a smaller displacement. The ratio of longest to shortest semi-axis is , the square root of the spectral condition number of a positive-definite matrix.[ref]
Strict positivity matters because gives
Hence a fixed positive level set cannot extend arbitrarily far. The Cholesky decomposition gives another view: , so the ellipsoid is the inverse image of a sphere under an invertible linear map. A singular positive semidefinite matrix has a nonzero kernel; adding a kernel vector preserves its quadratic form, so any nonempty positive level set is unbounded. These statements concern finite dimension; the positive definite symmetric operator entry explains why strict positivity alone need not give a uniform lower bound on an infinite-dimensional Hilbert space.
For a real symmetric matrix, or more generally a Hermitian matrix, the finite-dimensional spectral theorem makes all eigenvalues real. Intersecting its Gershgorin discs with the real axis gives
The union preserves gaps that the single outer interval loses. If every , all eigenvalues are positive, so the matrix is a positive-definite matrix. This is the positive-diagonal strictly diagonally dominant matrix criterion. For the three-by-three example, gives a positive-definite matrix certificate and bounds its spectral condition number of a positive-definite matrix by ; the computed value is about .
Strictness matters. The symmetric matrix has positive diagonal and is a diagonally dominant matrix, but its eigenvalues are and : it is singular. For a general complex matrix, the condition only puts the eigenvalues in the open right half-plane. The Hermitian matrix assumption is what converts that statement into positive definiteness.
Use the Euclidean norm and its induced operator norm. The spectral theorem for real symmetric matrices gives and . If , then the direct inverse has error , so
For nonzero , the relation also gives
This is the spectral condition number of a positive-definite matrix. The worst relative amplification is attained by exact data along an eigenvector for and noise along one for . A large therefore makes the reconstruction sensitive to relative data errors. Absolute sensitivity additionally depends on itself: scaling every eigenvalue equally leaves unchanged but changes . Since the matrix is finite dimensional and positive definite, the inverse exists and is continuous; the issue is ill-conditioning rather than mathematical nonexistence or discontinuity.
The linear-system residual at an approximate solution is . For symmetric positive definite , the conjugate gradient method has and minimizes the squared -norm of the error over that affine space; its residual is orthogonal to .
If has distinct eigenvalues, interpolate on them with a polynomial of degree at most . The finite-dimensional spectral theorem gives , so the exact correction lies in . The minimum possible error at step is therefore zero. Thus in exact arithmetic, possibly earlier if the initial residual has fewer eigencomponents. If , it terminates immediately.
A symmetric positive definite matrix preconditioner changes the system to , with ; practical preconditioned CG applies solves with . It can accelerate convergence by reducing the spectral condition number of a positive-definite matrix, which improves the usual bound involving , and by clustering eigenvalues, which allows low-degree residual polynomials small on the spectrum even when the extreme-eigenvalue bound is pessimistic.
Choosing makes the transformed matrix , so one step suffices. But applying already requires solving the original problem, and constructing an exact factorization generally incurs its original computational cost. It is therefore not a useful shortcut for a single unsolved system, though an available factorization can be reused for multiple right-hand sides. Effective matrix preconditioning balances spectral improvement against inexpensive application.
A particularly simple constructive choice uses paired collocation points for a global relation. For , put
Four suitable sets, one for each side, are
The braces in each set range over all . For large , compute to avoid cancellation. The corresponding pairs are, respectively,
Each pair has the same exponentially growing normal direction and opposite tangential frequencies. Take the phase-weighted difference of the two collocated global relations for a linear boundary value problem. For example, the bottom pair gives the adjoint test
The other three phase-weighted differences give
Thus the four spectral sets supply four real sine-tested equations per mode. We retain these equations; requiring every uncombined complex equation as well would instead form an overdetermined finite approximation. For real traces each retained equation is the appropriate real combination of its conjugate pair.
The key advantage is that vanish on the vertical sides, while vanish on the horizontal sides. Orthogonality then eliminates all off-mode unknowns. Define the known right-hand sides
The retained global relations for a linear boundary value problem are
For example, with , the bottom and top right-hand sides are explicitly
For completeness the left and right right-hand sides are
All these integrals use known polynomial approximations or the exact Dirichlet boundary condition.
Scale each row by . Every block becomes
There is just one off-diagonal entry in each row. This proves diagonal dominance of paired square global-relation collocation. Therefore the assembled system is strictly diagonally dominant, with row margin . This proof applies to the explicitly paired and scaled collocation system, rather than assuming arbitrary raw spectral rows are diagonally dominant. Each block is a positive-definite matrix and have spectral condition number of a positive-definite matrix
For instance, setting gives the concise answer
with the identical formula for the opposite vertical sides. If the known-data integrals use exact , these are exactly the first sine coefficients of the true normal traces: no discarded unknown mode contributes to a retained row. With a finite Legendre polynomial approximation, the error is solely that of the known-data approximation and the final normal-trace truncation. These square modified Helmholtz Dirichlet-to-Neumann coefficients determine the unknown boundary normal derivative; Green's third identity or the separated-variable construction can then recover the interior solution.
Because is a real symmetric positive-definite matrix, the finite-dimensional spectral theorem supplies an orthonormal eigenbasis with
Its eigendecomposition is also its singular value decomposition. If , then
so
The operator norms satisfy and . Therefore the worst-case relative perturbation bound is
The ratio is the spectral condition number of a positive-definite matrix. A large ratio means that data noise aligned with an eigenvector for the smallest eigenvalue is strongly amplified, so the inverse problem is ill conditioned.
For
the gradient is . The steepest descent method with exact line search for a positive-definite quadratic therefore uses
For and , the minimizer is . Put
Starting from , direct substitution gives and, inductively,
The Euclidean norm is unchanged by the alternating sign, so
For a positive-definite matrix, the spectral condition number of a positive-definite matrix is
Here , and hence
The conjugate gradient method starts with and updates
By finite termination of the conjugate gradient method, it reaches the exact solution in at most the number of distinct eigenvalues. This matrix has two, so at most
are required in exact arithmetic.
The linear-system residual is computable without knowing . The forward error is the actual difference from the desired solution. For an invertible matrix,
Keeping fixed, the approximate vector exactly solves with . Thus, for , is the relative backward error when only the right-hand side may change. For a vector norm and its induced operator norm,
A small relative linear-system residual ensures a small relative forward error when the condition number is moderate; an ill-conditioned matrix can amplify it. The bound follows directly from and .[ref]
For the Jacobi iteration for a three-variable tridiagonal system, the first linear-system residual is , while the first forward error is . In the Euclidean norm, the relative forward error is and the relative linear-system residual is . Its spectral condition number of a positive-definite matrix is , so the bound is about . The bound is a guarantee, not an equality. In floating-point arithmetic, recompute the linear-system residual from the current vector when checking termination; a small update alone need not certify a small forward error.
For a symmetric positive-definite matrix , solving is equivalent to minimizing the quadratic form with linear term
Writing the error vector and expanding gives the exact energy norm identity
The cross term vanishes because . With and , the same quantity is . Symmetric positive-definite preconditioning changes coordinates without changing the energy assigned to a corresponding error vector.
In an orthonormal eigenbasis of , the level set has semiaxis lengths . Its longest-to-shortest axis ratio is . A large spectral condition number of a positive-definite matrix therefore appears geometrically as a thin ellipsoid. Good scaling makes the transformed ellipsoid rounder, allowing a simple search direction to improve all coordinates together.
The negative gradient of is the linear-system residual . Relative to the norm , a steepest direction is : the Cauchy-Schwarz inequality applied to shows that the greatest decrease per unit -length occurs along this direction. Equivalently, ordinary gradient descent in becomes preconditioned steepest descent in .[ref]