Use the principal complex logarithm in the upper half-plane. Deform the interval upward into a half-strip: the vertical side at zero is traversed upward, and the side at one downward. The top side vanishes exponentially; a small indentation at zero contributes and vanishes. Thus the contour deformation gives the exact representation, for ,
Since , scaling and using the integral defining Euler's constant gives
For the second endpoint, . Watson's lemma, or integration of these powers against with an exponentially small tail, yields
In particular the required first terms are
The two different endpoint scales come from the logarithmic singularity at zero and the simple zero of the amplitude at one.
For , dominant balance for algebraic roots gives two roots on the scale and a third on the scale . On the first scale put . The equation becomes . The derivative of at either or is two, so the first correction is . On the second scale put ; then . Writing gives . Thus all three real-root expansions for positive are
These are fractional-power root expansions. To check that none is missing, the derivative of the polynomial is . Its nonzero critical points are ; for sufficiently small positive the polynomial is positive at the negative one and negative at the positive one. Its three monotone ranges therefore contain exactly these three real roots. The remaining two small roots correspond to the nonreal cube roots of unity in the second balance.
The printed limit does not specify the sign of . If a two-sided real limit is intended, also put with . The polynomial is strictly increasing and has just one real root. The same dominant balance for algebraic roots gives the negative-parameter branch
The Leibniz rule expands each derivative of a product into finitely many terms:
The symbol class estimates bound every summand by a constant times . Hence symbol orders add under multiplication:
We now address the remaining unheaded requests. To define the oscillatory integral as a functional on the space of test functions, choose equal to one near zero and set
The limit is not an assertion of absolute convergence of the original frequency integral. Split off a bounded-frequency part, which is smooth in . At large frequency define
Homogeneity and nonvanishing of the total phase differential imply on each compact spatial set. Also . The spatial coefficients of have symbol order , and its frequency coefficients have order zero; consequently its formal transpose of a differential operator lowers symbol order by one. After integrations by parts, the high-frequency pairing has an absolutely integrable amplitude , where is a fixed cutoff near zero. Derivatives of the outer cutoff yield errors bounded by times finitely many derivatives of , so they tend to zero. This proves cutoff independence of an oscillatory integral and defines a linear map into . Its distributional continuity is permitted as an assumption, and is also visible from this finite-derivative estimate.
The singular support of a distribution is the complement of the largest open subset on which it equals a smooth regular distribution. The stationary-direction bound for singular support needs conic support of an oscillatory amplitude. Write
Its radial lift is the closed conic enlargement of the ordinary support. The generally valid bound is
If the amplitude support is conic, this is exactly the printed bound. It is also the standard interpretation when support in the frequency directions is understood conically.
To prove the bound, take a point outside its right-hand side. Closedness of and compactness of the sphere give a neighborhood and with on the amplitude's directions over . On a slightly larger directional neighborhood use
The formal transpose of a differential operator lowers symbol order by one, using frequency derivatives only. An derivative of order of the oscillatory integrand has symbol order at most . Choosing more than integrations by parts makes that derivative absolutely integrable, uniformly on smaller compact subsets of . This works for every , so is smooth on .
The ordinary amplitude support can miss a singular-support limit phenomenon requires a qualification here: with unrestricted ordinary support the printed inclusion is false. An explicit counterexample uses , , one frequency variable and . Choose , and with . Its translates have disjoint supports. Choose an even smooth function that is zero for and one for , and put
This is a symbol of order zero. All spatial derivative bounds follow from ; frequency derivatives have the required decay because the th cutoff derivative is supported where . Near every point with finite , all large- terms vanish through the frequency cutoff and all remaining terms vanish in a small neighborhood. Thus no point lies in its ordinary support.
Nevertheless its oscillatory integral is
where is smooth. Indeed, its th term is the ordinary Fourier integral of times ; derivatives of order in and in are bounded by a constant times , a summable sequence. Since , each is singular. Closedness of singular support forces to be singular too, although the literal ordinary-support right-hand side omits it. The closed conic support includes this limiting point and resolves the defect.
Finally use Fourier inversion distributionally. The constant amplitude gives , and differentiation of the exponential supplies a factor . Thus the polynomial-amplitude integral is a delta derivative:
Its action on a test function is , confirming both the sign and the normalization. This is the delta derivatives from polynomial oscillatory amplitudes identity.
A phase function is a real function on , positively homogeneous of degree one in its frequency variable,
with nonvanishing total differential . Vanishing of its frequency gradient alone is allowed; those critical directions are relevant to singularities.
With , the symbol class consists of smooth functions such that, for every and every pair of multi-indices,
There is no requirement that the constant be uniform in the derivative indices. We use ; the factors of do not affect these estimates.
For , any further derivative satisfies
Thus frequency derivatives lower symbol order and spatial derivatives preserve it:
This is the differentiation rule in symbol calculus.
Let , a Schwartz function. Splitting the inverse Fourier transform at the jump in the Hilbert-transform Fourier multiplier gives
For , integration by parts on each half-line shows
Both restricted derivatives are in . The Riemann-Lebesgue lemma makes the bracket tend to zero as . Hence the two-sided tail is
Here . The large-distance tail of the Hilbert transform thus depends on the zeroth moment of the input. If that moment is nonzero, the tail proves that the output is not a Schwartz function, despite being smooth and square-integrable.
The Hilbert-transform Fourier multiplier is bounded, and is integrable for every nonnegative integer . Therefore its inverse Fourier transform can be differentiated under the integral arbitrarily many times:
These derivatives are continuous by dominated convergence theorem. Since , the Hilbert transform is smooth and commutes with differentiation:
The Fourier proof avoids differentiating a singular kernel without preserving its Cauchy principal value prescription.
We place the definitions and the unheaded preliminary requests here before addressing the first labelled property. The Schwartz space consists of smooth functions with finite seminorms
for every pair of multi-indices. Convergence means convergence in each seminorm. The tempered distribution space is its continuous dual space, with weak convergence tested against every Schwartz function. A continuous functional satisfies a bound by finitely many of these seminorms, equivalently by one sufficiently large weighted derivative seminorm.
Use the Fourier transform convention
Differentiation under the integral and integration by parts express as a constant of modulus one times the Fourier transform of . Its supremum is bounded by the norm of that function. For , this norm is at most a constant times finitely many Schwartz seminorms, using the integrable weight . Thus is continuous.
For completeness, Fourier inversion follows by inserting in the inverse integral and using Fubini's theorem. The result is convolution with the Gaussian approximate identity
It tends to , while integrability of allows the damping factor to be removed by dominated convergence theorem. Consequently . Reflection preserves every Schwartz seminorm, so the inverse transform is continuous as well. This proves the Fourier transform isomorphism of the Schwartz space.
Define the Fourier transform of a tempered distribution by transposition,
The Schwartz-space continuity just proved makes this a tempered distribution. Its inverse is , where . These maps are continuous for weak convergence, since each pairing is a pairing with a fixed transformed test. They are also continuous for the strong dual topology, because the Schwartz-space maps take bounded sets to bounded sets.
The convolution of a tempered distribution with a Schwartz function is
Smooth dependence of translated Schwartz functions gives . The finite-seminorm estimate and show that each derivative has at most polynomial growth. In particular, is a smooth function defining a tempered distribution. It need not itself be a Schwartz function; for example .
Writing , its distributional pairing is . The inner convolution is a Schwartz function, and this identity follows by integration in the Schwartz topology, justified by the weighted seminorm estimates. A direct Fubini's theorem calculation gives . Hence
Multiplication is well defined because multiplication by acts continuously on .
Now write the Hilbert transform as convolution with , the principal-value reciprocal distribution. The given Heaviside function transform, together with , yields
The value of the sign function at zero is irrelevant to its regular distribution. Thus the Hilbert-transform Fourier multiplier is
Applying Plancherel theorem, whose normalization here is , proves the isometry:
The principal-value integral agrees with this convolution: near its singular point subtract , and use odd cancellation; at infinity the Schwartz decay gives convergence. The Hilbert transform has domain , but generally does not take values in , as the tail in part (c) demonstrates.
Substitute part (b) into the spectral representation. The terms containing are
The transform is analytic for . Write . In the first quadrant,
which decays since . The Cauchy integral theorem in that quadrant equates the two integrals in the first parentheses. In the second quadrant,
which decays since , , and . The same contour integration equates the two integrals in the second parentheses. Therefore makes no contribution. This is upper-quadrant cancellation of reflected boundary transforms, and it uses the whole paired expression rather than attempting to discard an individual integral.
The terms involving cancel by the same two quadrant arguments with the analytic factors and omitted. There is a sign defect in the printed reconstruction prefactor. The sides specified in part (a) form a clockwise boundary, so the Cauchy integral theorem gives a negative reconstruction sign. Indeed, integrating each spectral ray first yields , and hence the sum of the three ray integrals is . Thus the printed positive prefactor reconstructs . The requested cancellation above holds with either overall sign, but the actual data-dependent derivative is
Decay at infinity fixes the integration constant when recovering from its Wirtinger derivative.
For completeness, the missing condition at infinity and the unrestricted printed deserve an explicit solvability of a decaying Robin strip check. Let be orthonormal eigenfunctions from a Sturm-Liouville problem of with and , and eigenvalues . With , separation of variables gives
Indeed, each coefficient obeys and . A positive eigenvalue has one decaying exponential; a zero eigenvalue has only affine solutions and a negative eigenvalue only oscillatory solutions, neither of which decays unless its coefficient vanishes. This also proves uniqueness in the decay class and justifies reflection symmetry there. For , all eigenvalues are positive. For , the constant eigenfunction requires ; the apparent zero of at the origin is then removable. For , compatibility with every nonpositive eigenmode is necessary. These restrictions cannot be inferred from smoothness and reflection symmetry alone. The printed boundary value problem without any far-field condition permits additional growing or nondecaying homogeneous solutions, whereas the spectral transforms select the decaying branch.
The Cauchy integral theorem applied to yields the global relation
Initially this follows for , where the far vertical closing segment vanishes. The identities below can be compared on the real axis, and then continued wherever the respective transforms are analytic.
Define the characteristic factor of the symmetric Robin strip transform
Substituting part (a) into the global relation gives
Reflection symmetry gives and , because . Also . Use the global relation at and multiply by to obtain
Comparison therefore proves
At a zero of , this identity is interpreted through the necessary solvability condition and a removable limit when a decaying solution exists. For , has no real zeros, so no such qualification is needed on the two real integration rays.
Use the Wirtinger derivative and write . Since solves the Laplace equation, is a holomorphic function. The transforms in the question select solutions with sufficient decay at infinity. In this class the solution, when it exists, is unique; hence reflection of the symmetric data about gives
For , this uniqueness follows directly from Green's first identity: the homogeneous problem has integral . For arbitrary real , the decaying-class qualifications are explained in part (c). Without a condition at infinity, symmetry of the data alone would not force symmetry of every solution.
Put , , and introduce a known transform
The bottom Robin boundary condition gives , while the top one gives . The bottom side is traversed from infinity to zero. Integration by parts therefore gives
The top side is traversed from towards infinity, and its exponential supplies , giving
On the vertical side, and the prescribed Neumann boundary condition is . Thus
Both the side orientations and the factor from the Wirtinger derivative are essential to these signs.
Splitting the free Schrodinger equation into real and imaginary parts gives
Differentiating the first relation in time and using the second yields the Euler-Bernoulli beam equation . This is the Schrodinger factorization of the elastic beam equation.
Assume the initial velocity has an integrable first spatial moment, as allowed by sufficient decay, and define
Then and decays at infinity. To encode the second boundary datum, define
The Dirichlet boundary condition for the resulting free Schrodinger equation is compatible at the corner, since . Insert these explicit into the data-only complex integral in part (b), with and defined as in part (a). The required displacement is the real part of that integral. Equivalently, the uniformly convergent lifted integral in part (b) may be used with the same complex data.
The Schrodinger factorization of the elastic beam equation verifies every condition: , , , and
The corner requirements on and ensure consistency of these derivative traces; the natural interpretation of the last printed compatibility is . If its prime were instead imposed for every , that would simply be an extra restriction on the data, and the same construction would still solve them.
The unknown Neumann boundary condition can be removed by the Fokas method. Evaluate the global relation for the half-line free Schrodinger equation at :
Consequently . In the representation from part (a), the last term has integral , by the Cauchy integral theorem and Jordan lemma. It is analytic in the upper half-plane, and the exponential decays on the closing first-quadrant arc. Hence an expression involving only the given data is
One may replace by in the upper limit defining : the contribution of boundary times closes to zero in , since then decays there. Using makes causality transparent.
For an explicit proof of uniform convergence, it is useful to apply boundary lifting before inversion. Set , , , and
The compatibility condition gives . The Fourier sine transform of satisfies , with initial value . The integrating factor therefore gives the equivalent representation
This last integral is absolutely and uniformly convergent for , under concrete sufficient hypotheses , decay of the boundary terms, and . Indeed, two integrations by parts give for . Another integration by parts, this time in , gives
Thus the initial term is uniformly and the forcing term uniformly . For , use and the bounded time integral. An integrable majorant proves the claimed uniform convergence and permits evaluation at both boundaries.
At , Fourier sine inversion gives . At , the integral vanishes, giving , including the compatible corner. To verify the equation, note that and the transformed equation implies
Adding the lifted part gives . Under the stated smoothness, this holds classically in the interior; differentiated spectral integrals can first be Gaussian-regularized, or read in the sine-transform sense and then identified with the smooth solution. Uniform convergence of itself does not require claiming uniform convergence of every differentiated integral at the corner.
Use the half-line Fourier transform and finite-time boundary transforms
The half-line Fourier transform is analytic for under spatial decay. All spectral integrals below have their usual oscillatory, or vanishing Gaussian damping, interpretation until absolute convergence is established.
The free Schrodinger equation has the local divergence identity
Integrating in gives the global relation for the half-line free Schrodinger equation
Let . Orient its boundary from down to , and then from to , so that lies on the left. Fourier inversion followed by contour integration in the second quadrant yields
This is a complex spectral representation involving the initial trace and both boundary traces. The contour deformation works because each boundary-time integrand contains with , which decays in the second quadrant, as well as for . This fixes both the quadrant and the orientation; changing either without changing the signs would produce an incorrect representation.
Fix . If , all knot values of the function agree with a polynomial of degree . Its order- divided difference vanishes. If , all the knot values are zero, so the same divided difference again vanishes. Thus the support is finite:
The support and Gram bandwidth of B-splines also holds at repeated knots: the same conclusion follows with the standard limiting convention. If , then , so the two supports have no positive-length overlap. An endpoint intersection contributes zero to the integral defining . The case is identical. Hence
For strictly increasing knots, B-splines whose indices differ by overlap on a positive-length interval and their product is positive there, so this universal bound is sharp. In standard matrix terminology the half-bandwidth is and the full bandwidth is ; the paper's integer is .
Use the standard partition normalization of B-splines, not rescaling each spline to have supremum exactly one. Let . Then . In the ordinary real inner product, the mixed-normalization spline Gram matrix is , where and . Since the form a basis, is positive definite; is invertible, so is invertible. The undeclared upper index in the printed matrix should be .
We need the standard positivity and normalization properties with their justification. The Cox-de Boor recurrence is
with zero-width terms interpreted as zero. It starts from nonnegative interval indicators. On each lower-order spline's support, its recurrence coefficients are nonnegative, so induction gives . Extend the finite knots outside their endpoints to a full knot sequence. In the full sum, coefficients of each lower-order spline add to one, so the recurrence preserves partition of unity. Our finite collection is a subset of nonnegative B-splines; therefore everywhere. The Marsden identity gives equality on the basic knot interval.
The divided-difference definition also gives support in : below the left endpoint the knot-data function is a polynomial of degree , whose order- divided difference vanishes; above the right endpoint the knot data vanish. Choose and . Integrating in and then taking the divided difference in its knot variable gives
because an order- divided difference of a monic degree- polynomial is one. The support lies inside , so .
Orthogonality of to the spline space gives , with . Since and has unit integral, . The induced matrix maximum norm is the absolute row-sum norm, so
Taking the supremum over nonzero proves the maximum-norm bound for spline projection:
The range is , so the printed phrase “onto ” must mean into , and onto . Continuity of the range requires continuous B-splines, for example and internal multiplicities at most . For order-one or discontinuous B-splines, the same bound is valid with codomain rather than .
Fix the degree bound . Existence can be proved without a general uniqueness theorem: a minimizing sequence in is uniformly bounded because its distances from are bounded. Values at any fixed distinct points determine its coefficients through an invertible Vandermonde matrix, so those coefficients are bounded. A convergent subsequence supplies a polynomial attaining the minimum.
For uniqueness of best uniform polynomial approximation, suppose and attain the same minimum , and let . The triangle inequality makes another minimizer. If , both polynomials equal , so assume . Put and . At each , the two real errors and lie in and their average equals an endpoint. Hence they are equal there, and .
The set must contain at least points. Otherwise interpolate the values on its at most points by a polynomial . Then on . Continuity gives this positivity on a neighbourhood of , while the error has a strict gap below on the compact complement. A sufficiently small positive multiple of decreases the maximum error, contradicting optimality of . This is an elementary perturbation argument, not an appeal to Haar's theorem.
Thus has at least distinct zeros. A polynomial of degree at most with that many zeros is identically zero. The best uniform polynomial is unique for every fixed degree bound: .
Let and , with . The Chebyshev alternation theorem is
If such points exist and a polynomial had at every extremum, its signs at those points would alternate. The intermediate value theorem would give at least distinct zeros, impossible for a nonzero degree-at-most- polynomial. The Kolmogorov criterion for uniform approximation therefore proves sufficiency.
For necessity, let and . They are disjoint compact sets. If one set is empty there is just one sign block. Otherwise their positive separation implies that, reading the extremal set from left to right, its signs form finitely many alternating blocks. If there are fewer than alternating extrema, there are at most blocks. Place one point in each extremum-free gap between consecutive blocks. A scalar multiple of can be chosen to have exactly the sign of on every block. Its degree is at most , and throughout , contradicting the Kolmogorov criterion for uniform approximation. Thus there must be at least alternating extrema. If , the exact polynomial is already best and the equalities with zero error are automatic.
Let , , , and . The real Kolmogorov criterion for uniform approximation says is best if and only if
Equivalently, no direction has the error's sign strictly at every extremal point. If , the criterion is automatic.
For sufficiency, a point with gives , so cannot improve the norm. For necessity, suppose on . Compactness gives a positive lower bound there and on a neighbourhood of . On that neighbourhood, for sufficiently small positive . Off the neighbourhood, has a strict gap below , and small preserves that gap. Thus has smaller error, a contradiction. This also explains why the criterion involves all active extrema.

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