For the growing normal mode in two deep layers,
It vanishes at the endpoints of the unstable band and has its unique interior maximum where . Using gives
This is the maximum exponential growth rate, not the real oscillation frequency.
Put . In the deep-layer limit, the finite-depth Rayleigh-Taylor dispersion relation gives . Its sign changes at
The Rayleigh-Taylor instability band is ; is neutral and is oscillatory.
Use the linearized boundary condition: evaluate at and discard products of perturbations. The kinematic boundary condition and Young–Laplace equation become
Incompressible flow and irrotational flow give the Laplace equation for each potential. For a normal mode, the wall conditions select
With , the kinematic conditions give and . Hence and . Substitution into the dynamic condition gives the finite-depth Rayleigh-Taylor dispersion relation:
Its denominator is positive. Assuming and , negative therefore produces a growing branch with , as in Rayleigh-Taylor instability.
The interface is material: fluid particles on remain there. Its material derivative therefore vanishes, giving on the interface. Since , this is the kinematic boundary condition for a free-surface graph on each side. The material derivatives use the respective tangential fluid velocities; they enforce the same moving normal boundary.
The Unsteady Bernoulli equation gives . Absorb the spatially constant functions into the gauges of the velocity potentials. Evaluating at and subtracting the equations yields the stated dynamic boundary condition for an inviscid interface. Surface tension supplies the pressure jump, rather than pressure continuity. Thus the first condition enforces material motion, and the second enforces the Bernoulli pressure balance.
For irrotational flow, . An impermeable stationary horizontal wall has zero normal velocity, so the kinematic boundary condition is
No tangential no-slip condition is imposed on these inviscid fluids.
Write for the indicator functional of a constraint set and . Then
Both terms are convex, hence their infimal convolution is the convex squared distance to a convex set. The closest point theorem in a Hilbert space supplies the unique metric projection onto a closed convex set and gives .
Identify the Hilbert space with its dual space using the Riesz representation theorem. Completing the square yields , while is the support function. The conjugate of the squared distance to a convex set is therefore
For the closed unit ball, the metric projection onto a closed convex set is if , and otherwise. Therefore
The last equality uses the Cauchy-Schwarz inequality to compute the unit ball's support function, attained in the direction of when .
For , the definition of the convex conjugate gives
where the substitution is a bijection of the independent pairs. Therefore
This infimal convolution identity does not need convexity of or attainment of the inner infimum. The assumptions ensure that both functions have nonempty effective domains; their conjugates never take , so the separated sum is well defined, allowing .
Let . To prove convexity, take with finite values, , and . By the definition of infimum, choose with
Put and . Convexity of and gives
Let . If either endpoint value is infinite, the convexity inequality is automatic, and gives equality. Thus the infimal convolution of two convex functions is convex under the stated properness assumption. The proof uses approximate minimizing splits, so no attainment of the inner infimum is assumed.
For a functional on a real Banach space, its convex conjugate is defined on the dual space by
Here is the dual pairing. For complex spaces, use its real part, equivalently regard the space as real for convex analysis. The convex conjugate is a convex function, being the supremum of affine functions of , even when itself is not convex.
The small-parameter limit depends on data compatibility with the regularizer domain. Set
For every such , optimality gives
Taking the upper limit as and then the infimum over proves . Its two nonnegative contributions above must vanish, yielding the general result
In particular, the requested zero-misfit limit holds when . A sufficient condition is an exact solution with and ; then
Membership of in alone is insufficient. Take , , , and the indicator functional of a constraint set . All positive-parameter objectives are proper extended-real functions with coercivity and sequential lower semicontinuity, with the unique minimizer , yet for every . Thus the first printed limit needs compatibility with the regulariser's effective domain; the second limit remains true under the given assumptions.
Comparison with zero and nonnegativity of the two terms gives
Therefore
For a concrete variational regularization example, take a Hilbert space and . The bound forces in norm. Since is a bounded linear operator, , and continuity of the squared norm gives
The same conclusion is not forced by for a regulariser with a nontrivial zero set.
For , abbreviate and . Optimality of the two variational regularization solutions gives
Adding shows , so . Rearranging each comparison then gives the monotonicity of data fidelity and regularization penalty
Hence the regulariser value is non-increasing and the data misfit is non-decreasing with the parameter. These inequalities hold for any choices of minimizers at the two distinct parameters, including when uniqueness fails.
Use the value function of variational regularization in its selection-independent form
Because , each objective is non-decreasing with , hence so is its infimum. For , the affine function of the parameter at each fixed gives
Thus is concave.
For continuity, on any interval , comparison using gives
It follows that is a locally Lipschitz function, hence continuous, on . Therefore the value function of variational regularization is non-decreasing, concave and continuous. No uniqueness of minimizers is needed for these conclusions.
First interpret the stated well-definedness as including uniqueness of the global minimizer for each positive parameter. Under that assumption, the parameter continuity of variational regularization follows from the direct method in the calculus of variations argument below.
Let . For large , . Comparison with zero, using , gives
Coercivity of therefore bounds the sequence. Take a -convergent subsequence with limit . For any , optimality gives
The last two terms vanish. Sequential lower semicontinuity at the fixed parameter implies . Thus every cluster point is a global minimizer at . Uniqueness makes it , and the subsequence property proves
Indeed, a subsequence staying outside a neighborhood of would have a further convergent subsequence, whose limit must be , a contradiction.
The listed existence hypotheses alone do not imply uniqueness or convergence of an arbitrary selection. For example, on , set , , and
This is nonnegative, convex, continuous and has ; every has coercivity. Its minimizers are for all . Along , selecting prevents convergence. Without uniqueness, the valid conclusion is that every cluster point minimizes the limiting objective.
The Iterated Tikhonov regularization update is the proximal operator step
The objective has coercivity and is weakly sequentially lower semicontinuous and strongly convex, so the direct method in the calculus of variations and uniqueness of a minimizer of a strictly convex function give existence and uniqueness. Its gradient equation is
which is exactly . The second term penalizes the change from the previous iterate, rather than the norm of relative to zero.
Assume , as in the exact-solution setting used for the preceding source condition for quadratic regularization. Let , , and . Then , and expanding the two squared errors gives
The Cauchy-Schwarz inequality supplies the last line. Thus while the residual remains at least the noise bound,
Compatibility of the exact data is essential. If only is retained, let , , , and . The minimum-norm least-squares solution is zero, but the first iterate is and its residual norm is . Its error has increased from zero. Thus the original assertion uses the implicit exact-data assumption from part (c), and is false without it.
Write the residual of an inverse problem as . The iteration can be rearranged as
Applying gives . Since is a positive operator, its resolvent operator has operator norm at most one. Hence
More explicitly, taking the inner product of the residual equation with gives
This directly proves the same monotonicity by the Cauchy-Schwarz inequality, including the case . An unresolvable data component in remains unchanged.
With exact data , the iteration error obeys
The source condition for quadratic regularization excludes any null space component. Using its singular system of a compact operator coefficients,
By Bernoulli's inequality, . Also . Therefore
Summing the squared coefficients and using Bessel inequality yields
This is an a priori error rate from the source condition for quadratic regularization, with constant independent of .
Let , so . By linearity, . The preceding operator norm estimate gives
Moreover, has factors on the data singular vectors, and is zero on . Every factor is in , hence
Thus, for any fixed , one can take and .
The first constant cannot generally be chosen independently of the iteration count. On the l2 sequence space, take , , and noise . At the solution perturbation has norm
This disproves a uniform-in- interpretation of the stated first estimate. For a closed operator range, boundedness of the Moore–Penrose inverse of an operator instead supplies the uniform constant .
For Iterated Tikhonov regularization, set and . In a singular system of a compact operator, each coefficient obeys
Summing this geometric recurrence gives
Equivalently, its iterated Tikhonov spectral filter is . For , the finite geometric sum gives
Consequently the inverse coefficient is at most . Thus every is a bounded linear operator with
For exact admissible data, write . The Picard criterion gives , and
by dominated convergence theorem. Therefore this is a linear regularization of an inverse problem with discrete parameter ; a family for all positive can be obtained by setting . No upper restriction on the positive step size is needed for this implicit iteration.

Pinned article: Introduction to the OurBigBook Project

Welcome to the OurBigBook Project! Our goal is to create the perfect publishing platform for STEM subjects, and get university-level students to write the best free STEM tutorials ever.
Everyone is welcome to create an account and play with the site: ourbigbook.com/go/register. We belive that students themselves can write amazing tutorials, but teachers are welcome too. You can write about anything you want, it doesn't have to be STEM or even educational. Silly test content is very welcome and you won't be penalized in any way. Just keep it legal!
We have two killer features:
  1. topics: topics group articles by different users with the same title, e.g. here is the topic for the "Fundamental Theorem of Calculus" ourbigbook.com/go/topic/fundamental-theorem-of-calculus
    Articles of different users are sorted by upvote within each article page. This feature is a bit like:
    • a Wikipedia where each user can have their own version of each article
    • a Q&A website like Stack Overflow, where multiple people can give their views on a given topic, and the best ones are sorted by upvote. Except you don't need to wait for someone to ask first, and any topic goes, no matter how narrow or broad
    This feature makes it possible for readers to find better explanations of any topic created by other writers. And it allows writers to create an explanation in a place that readers might actually find it.
    Figure 1.
    Screenshot of the "Derivative" topic page
    . View it live at: ourbigbook.com/go/topic/derivative
  2. local editing: you can store all your personal knowledge base content locally in a plaintext markup format that can be edited locally and published either:
    This way you can be sure that even if OurBigBook.com were to go down one day (which we have no plans to do as it is quite cheap to host!), your content will still be perfectly readable as a static site.
    Figure 5. . You can also edit articles on the Web editor without installing anything locally.
    Video 3.
    Edit locally and publish demo
    . Source. This shows editing OurBigBook Markup and publishing it using the Visual Studio Code extension.
  3. https://raw.githubusercontent.com/ourbigbook/ourbigbook-media/master/feature/x/hilbert-space-arrow.png
  4. Infinitely deep tables of contents:
    Figure 6.
    Dynamic article tree with infinitely deep table of contents
    .
    Descendant pages can also show up as toplevel e.g.: ourbigbook.com/cirosantilli/chordate-subclade
All our software is open source and hosted at: github.com/ourbigbook/ourbigbook
Further documentation can be found at: docs.ourbigbook.com
Feel free to reach our to us for any help or suggestions: docs.ourbigbook.com/#contact