Fourier analysis

Spatial frequency is the number of oscillation cycles per unit length. A ripple with spatial period $\Lambda$ has frequency $\nu =1/\Lambda$; its angular wavenumber is $k=2\pi \nu$. In two dimensions the spatial-frequency vector specifies the orientation and period of a plane-wave pattern.

A suitably band-limited function can be reconstructed from equally spaced samples when the sampling rate exceeds twice its largest ordinary frequency. Sampling more slowly aliases distinct frequencies in the Fourier transform. For spatial sampling, the same statement relates sample pitch to the shortest resolvable wavelength.

When the continuous Fourier transform $F$ of an integrable function vanishes outside $[-\pi ,\pi ]$, its periodic Fourier coefficients are the integer samples of $f$, with the sign of the index reversed. Projecting $F$ in $L^2[-\pi ,\pi ]$ and then applying the Fourier inversion theorem gives the displayed sinc function reconstruction. The Cauchy-Schwarz inequality controls the reconstruction error uniformly in $t$. Bessel's inequality bounds the shifted sinc coefficient vector, giving absolute uniform tails when the sample sequence belongs to ${\ell }^2$.

Samples spaced by pitch $p$ can distinguish spatial frequencies strictly below $1/(2p)$ cycles per unit length. Thus the shortest limiting wavelength is $2p$; exactly at the limit a sinusoid's phase can make all samples vanish. A practical deformable mirror has additional limits from actuator influence functions and finite pupil geometry.

A Euclidean lattice is a discrete subgroup $\Lambda \subseteq {\mathbb{R}}^n$ spanning the ambient vector space. Its covolume $m(\Lambda )$ is the volume of a fundamental parallelepiped.

For the skew matrix $S=\left(\begin{array}{cccc}0& 1& 1& 1\\ -1& 0& -1& 1\\ -1& 1& 0& -1\\ -1& -1& 1& 0\end{array}\right)$, both Euclidean lattices reduce to the tetracode modulo $3$. A reflection depending on the reduction coset matches their vectors coordinatewise in absolute value. Their weighted length spectra agree for all positive diagonal axis weights. Rationally independent weights force any possible linear isometry to be a coordinate sign change; the code then forces an overall sign, which does not identify the two lattices.

The covolume of a full Euclidean lattice is the Lebesgue measure of a fundamental parallelepiped in its ambient inner-product space. In orthonormal coordinates it is $|\det B|$ for a lattice basis matrix $B$. Scaling the lattice by $t>0$ multiplies the covolume by $t^n$; the dual lattice has reciprocal covolume.

In a rank-two Euclidean lattice, a shortest nonzero vector $v$ and a shortest vector $w\notin \mathbb{Z}v$ form a basis. The shortest vector is primitive. If the vertical coordinate of $w$ has index $m\ge 2$ in the projection lattice, a vector with smaller positive vertical coordinate can be reduced horizontally to length squared at most $(|v{|}^2+|w{|}^2)/4<|w{|}^2$, a contradiction. This lemma need not hold in higher rank.

For a full-rank Euclidean lattice of covolume $A$ in ${\mathbb{R}}^d$, bounded fundamental cells compare lattice points in a radius-$R$ ball with the volumes of balls of radii $R\pm C$. Hence $N(R)=\mathrm{vol}(B_1)R^d/A+O(R^{d-1})$. A lattice norm multiset therefore determines the covolume, which is useful for the spectrum of a flat torus.

The dual lattice isIts covolume is $m(\Lambda {)}^{-1}$.

Every continuous group homomorphism ${\mathbb{R}}^n/\Lambda \to {\mathbb{C}}^{\times }$ is $x+\Lambda \mapsto e^{2\pi i⟨u,x⟩}$ for a unique $u$ in the dual lattice. Compactness forces unit modulus, and lifting the character to the simply connected cover makes its phase a real linear functional. These characters are the frequencies of torus Fourier series.

For a sufficiently rapidly decreasing function $f$, the Poisson summation formula relates its samples to those of its Fourier transform:under the convention $\hat{f}(\xi )={\int }_{\mathbb{R}}f(x)e^{-ix\xi }dx$.

For a Schwartz function on the real numbers, its periodization is a smooth periodic function of period one. Rapid decay gives uniform convergence on $[0,1]$ of the series and each differentiated series. Under the Fourier transform convention $\hat{f}(\xi )=\int f(x)e^{-2\pi ix\xi }dx$, its $m$th Fourier coefficient is $\hat{f}(m)$: integrate termwise over one period and combine the translated intervals. The rapidly decreasing coefficients give a Fourier series with uniform convergence. Evaluation at zero proves the Poisson summation formula.

For a Schwartz function under the convention $\hat{f}(y)={\int }_{{\mathbb{R}}^n}f(x)e^{-2\pi i⟨x,y⟩},dx$,

For integers $m$ and $n$,This orthogonality extracts matching coefficients when two Fourier series are integrated over one period.

For the trigonometric polynomial $F(t)={\sum }_{k=1}^N{b}_{k}e(kt)$, orthogonality of complex exponentials givesIndeed, expand each square and use ${\int }_0^{1}e(jt)dt=0$ for every nonzero integer $j$, and one for $j=0$.

A wavelet is a localized function whose translates and dilates analyze different positions and scales.

From a one-dimensional multiresolution analysis, the simultaneous-scale two-dimensional detail space splits as $(W_j\otimes V_j)\oplus (V_j\otimes W_j)\oplus (W_j\otimes W_j)$. Its generating wavelets are $\psi \otimes \phi$, $\phi \otimes \psi$, and $\psi \otimes \psi$. Their normalized dilates are $2^j\Psi (2^j\cdot -k)$ for $k\in {\mathbb{Z}}^2$. Include a coarse scaling family when working on a bounded square.

A finite-length smooth curve meets $O(2^j)$ supports of localized two-dimensional wavelets at scale $j$. For a bounded function, their coefficients are $O(2^{-j})$, so their total squared energy at that scale is $O(2^{-j})$. Keeping these coefficients through level $J$ uses $O(2^J)$ terms and leaves squared error $O(2^{-J})$. Polynomial cancellation, or adequate approximation of the remaining smooth regions, therefore gives best N-term approximation squared error $O(N^{-1})$.

A tensor-product wavelet is a product of one-dimensional wavelets and scaling functions, with at least one wavelet factor. In two dimensions, the three simultaneous-scale types are $\psi \otimes \phi$, $\phi \otimes \psi$ and $\psi \otimes \psi$. Their translates and normalized dilates form a tensor-product wavelet basis.

An interval-adapted wavelet basis is an orthonormal basis of $L^2[0,1]$ made of a finite coarse scaling function family and resolution-indexed wavelets. A localized construction modifies only a bounded number of functions near each endpoint at every level. Boundary modification must preserve nested refinement spaces and the required vanishing moments; simple restriction of a whole-line basis does not do so.

A boundary wavelet replaces a translated interior wavelet whose support of a function meets a domain endpoint. Compatible finite boundary refinement matrices preserve orthogonality and localization across scales. If the coarse boundary spaces reproduce polynomials of degree less than $q$, their detail-space orthogonal complements have $q$ vanishing moments.

This construction adapts compactly supported orthonormal wavelets to an interval by finite changes near its endpoints. It preserves local support, polynomial reproduction, nested approximation spaces and orthogonal detail spaces. The resulting basis contains coarse scaling functions, unchanged interior wavelets, and finitely many boundary wavelets per endpoint and per scale, without forcing periodic or zero boundary values.

A wavelet has $q$ vanishing moments when these integrals vanish for $0\le r<q$. For compact support, this is equivalent to a zero of order at least $q$ at the origin of its Fourier transform. Such a wavelet annihilates polynomials of degree less than $q$, and a Taylor polynomial bounds coefficients on smooth regions.

Suppose an integrable orthonormal scaling function has a low-pass filter of a multiresolution analysis that is $C^{p-1}$ near $\pi$, and its Fourier transform is $C^{p-1}$ near zero. If the associated wavelet has $p$ integrable vanishing moments, then $m^{(k)}(\pi )=0$ for $k<p$. Indeed, moment differentiation of the Fourier transform makes ${\hat{\psi }}^{(k)}(0)=0$, while $|\hat{\phi }(0)|=1$. In $\hat{\psi }(2t)=e^{-it}\overline{m(t+\pi )}\hat{\phi }(t)$, division by the nonzero smooth factor proves the conclusion. With only continuity of that factor one still obtains a Peano zero, but not automatically higher ordinary derivatives.

Assume the scaling function Fourier transform and the MRA low-pass filter are $C^{p-1}$, and $m^{(k)}(\pi )=0$ for $k<p$. For every nonzero integer $j$, write $j=2^r\ell$ with $\ell$ odd. Iterating the scaling refinement equation at $2\pi j+t$ supplies a factor $m(\pi \ell +t/2^{r+1})$. All its derivatives of order less than $p$ vanish at $t=0$, by periodicity. The product rule gives ${\hat{\phi }}^{(k)}(2\pi j)=0$ for $k<p$.

An orthonormal wavelet $\psi$ generates $\{2^{j/2}\psi (2^{j}x-k){\}}_{j,k\in \mathbb{Z}}$ as an orthonormal basis of $L^2(\mathbb{R})$.

A Daubechies wavelet of order $q$ is a compact support orthonormal wavelet with $q$ vanishing moments and a minimal-length finite refinement filter. The minimal filter has $2q$ taps, giving support length $2q-1$ in the standard dyadic normalization. Its low-pass symbol has a zero of order $q$ at $\pi$. Increasing order improves polynomial cancellation and, for sufficiently large order, regularity, at the price of a wider support.

The Haar wavelet has one vanishing moment and compact support of unit length. Its scaling function is ${\chi }_{[0,1)}$, and its refinement coefficients are $h_0=h_1=1/\sqrt{2}$. The approximation spaces consist of functions constant on dyadic cells. Its discontinuity makes it effective for jump localization but limits smooth-function approximation.

The normalized Haar scaling functions and wavelets obeyThis orthogonal two-by-two transformation gives the multiresolution analysis decomposition $V_{j+1}=V_j\oplus W_j$. Iterating it expresses a Haar approximation either through level-$j$ cell averages or through coarse averages and all wavelet details at lower levels. The identities remain valid locally for coefficient integrals of locally integrable functions, without a global $L^2$ assumption.

The Haar projection at level $J$ is the orthogonal projection onto the functions constant on level-$J$ dyadic intervals. It replaces a function by its mean on each cell. It can be expressed using level-$J$ Haar scaling functions, or using the constant function and Haar wavelets at levels below $J$.

Finite total variation of a function on $\mathbb{R}$ gives the following error bound for averaging on dyadic cells of length $\delta =2^{-j}$:The cell error is at most its length times the oscillation on that cell, and the sum of the oscillations is at most the total variation. This proves that the difference is integrable even if $f$ is not globally in $L^1$. For an even function decreasing to zero on the positive half-line, $\mathrm{TV}(f)\le 2f(0)$ and the bound becomes $f(0)2^{1-j}$.

In the Gaussian white noise model, estimate a Haar scaling function coefficient by $\int {\varphi }_{J,k}dY$. The resulting Haar projection estimator is an unbiased estimator with independent coefficient errors of variance $1/n$. On each cell it is the observed path increment divided by the cell length. This gives a finite-dimensional estimator without imposing smoothness on the drift.

The cellwise error of the Haar projection estimator in Gaussian white noise is $\sqrt{2^J/n}$ times a standard normal variable. Its supremum norm is therefore exactly that factor times the Gaussian maximum over $2^J$ cells. The Gaussian maximum bound without independence gives expected error at most $\sqrt{2^J(2J+2)\log 2/n}$. This concerns the stochastic error about the projection; approximation bias must be added when estimating the full drift.

At level $J$, a Haar scaling function is the normalized indicator function $2^{J/2}{1}_{I_{J,k}}$ of a dyadic interval. The functions at that level form an orthonormal basis of the cellwise-constant subspace. Their normalization is what turns white-noise coefficient variances into $1/n$ and cellwise variances into $2^J/n$.

A multiresolution analysis is a nested sequence of approximation spaces related by dyadic dilation and generated at one scale by translates of a scaling function.

A scaling function generates the approximation space $V_0$ of a multiresolution analysis by its integer translates. In an orthonormal construction those translates form an orthonormal basis, and ${\phi }_{j,k}(x)=2^{j/2}\phi (2^{j}x-k)$ generate $V_j$. Integrable orthonormal scaling functions have $|\hat{\phi }(0)|=1$, as follows from density and the MRA projection Fourier identity.

With $\hat{\varphi }(\xi )=\int \varphi (x)e^{-ix\xi }dx$ in the Plancherel theorem sense, the integer translates of $\varphi \in L^2(\mathbb{R})$ are orthonormal exactly whenThe Fourier coefficients of the periodized energy are the translate inner products. Uniqueness of Fourier coefficients in L1 therefore proves both directions. No absolute integrability assumption on $\varphi$ is required.

If a periodic unimodular function $a$ and its inverse have absolutely summable Fourier coefficients, replacing ${\hat{\phi }}_0$ by $\hat{\phi }=a{\hat{\phi }}_0$ preserves the orthonormal integer-translate span and integrability of an integrable scaling function. In physical space this is an absolutely summable combination of integer function translations. The new MRA low-pass filter is $m(t)=a(2t)m_0(t)/a(t)$. A rough phase can therefore preserve the multiresolution analysis while changing MRA low-pass filter regularity.

LetThe Fourier coefficients of $\theta$ are absolutely summable, so $a$ belongs to the Wiener algebra, as does $a^{-1}$. The identity $\theta (2t)=\theta (t)+\theta (t+\pi )$ implies $a(2t)=a(t)a(t+\pi )$. Starting with a compactly supported Daubechies wavelet having $p\ge 3$ vanishing moments, the periodic phase change of a scaling function gives $m(t)=a(t+\pi )m_0(t)$, while the canonical high-pass construction leaves $\hat{\psi }(2t)=e^{-it}\overline{m_0(t+\pi )}{\hat{\phi }}_0(t)$ unchanged.

At a dyadic point $t_0=2\pi k/2^j$, the terms of the difference quotient with $n\ge j$ and $2^n|h|\le 1$ each contribute $1+O((2^{n}h{)}^2)$. Their number tends to infinity, their total error is bounded, and the remaining tail contributes a bounded amount. Hence $(\theta (t_0+h)-\theta (t_0))/h={\log }_2(1/|h|)+O(1)$, so neither $\theta$ nor $a$ has a finite derivative there. These points are dense. Away from the isolated zero of $m_0$ near $\pi$, multiplication by its nonzero smooth value cannot remove this nondifferentiability. Thus the new MRA low-pass filter is not differentiable throughout any neighborhood of $\pi$, despite integrability of the scaling function and unchanged vanishing moments. Ordinary higher derivatives at $\pi$ cannot be inferred, although the corresponding Peano zero survives.

Nesting in an orthonormal multiresolution analysis expresses a scaling function in the finer-scale orthonormal basis. The associated low-pass symbol is $m(\xi )=2^{-1/2}{\sum }_k{h}_k{e}^{-ik\xi }$, and $\hat{\phi }(2\xi )=m(\xi )\hat{\phi }(\xi )$. If $\phi$ has compact support, only finitely many coefficients $h_k$ are nonzero. With $\hat{\phi }(0)=1$, iteration gives $\hat{\phi }(\xi )={\prod }_{j\ge 1}m(\xi /2^j)$ when the limit exists.

For an orthonormal scaling function satisfying $\phi (x)=\sqrt{2}{\sum }_k{h}_k\phi (2x-k)$, its low-pass Fourier series symbol is $m(\xi )=2^{-1/2}{\sum }_k{h}_k{e}^{-ik\xi }$. The scaling refinement equation becomes $\hat{\phi }(2\xi )=m(\xi )\hat{\phi }(\xi )$. The symbol is initially an almost-everywhere defined periodic function; the assumption that the scaling function is Lebesgue integrable alone does not make it smooth. The orthonormal translates give the quadrature mirror filter identity.

For an orthonormal wavelet refinement filter, ${\sum }_k{h}_k\overline{h_{k-2n}}={\delta }_{n0}$ is equivalent to the displayed identity. A high-pass filter can be chosen with coefficients $g_k=(-1{)}^k\overline{h_{1-k}}$. In frequency, its symbol is $e^{-i\xi }\overline{m(\xi +\pi )}$, up to a constant phase. The two channels split an approximation space into a coarser approximation and its orthogonal complement.

For a scaling refinement equation $\varphi ={\sum }_n{a}_n\varphi (2\cdot -n)$, set $m(t)=\frac{1}{2}{\sum }_n{a}_n{e}^{-int}$. The quadrature mirror filter identity makes the two rows $(m(t),m(t+\pi ))$ and $(q(t),q(t+\pi ))$ a unitary matrix, where $q(t)=-e^{-it}\overline{m(t+\pi )}$. The complementary coefficients $b_n=(-1{)}^n\overline{a_{1-n}}$ yield $\psi ={\sum }_n{b}_n\varphi (2\cdot -n)$. In a multiresolution analysis, coefficient-space completeness shows its translates form an orthonormal basis of $V_1\ominus V_0$, hence its dilations and translates form an orthonormal wavelet basis of the whole space.

Use $\hat{f}(\xi )=\int f(x)e^{-ix\xi }dx$ and let $P_j$ be the orthogonal projection onto the closed span of the orthonormal translates and dilates of a scaling function. If $\hat{f}$ is supported in $[-R,R]$ and $R<\pi 2^j$, thenIndeed, the Plancherel theorem writes the coefficient against ${\phi }_{j,k}$ as $2^{-j/2}(2\pi {)}^{-1}\int \hat{f}(\xi )\overline{\hat{\phi }(2^{-j}\xi )}{e}^{ik{2}^{-j}\xi }d\xi$. With $\xi =2^j\theta$, these are $2^{j/2}$ times the Fourier series coefficients of $\hat{f}(2^j\theta )\overline{\hat{\phi }(\theta )}$ on $[-\pi ,\pi ]$. The Parseval identity proves the formula. It shows that continuity and unit modulus at zero imply density of the refinement spaces, and conversely that density forces this unit modulus when the Fourier transform is continuous at zero.

The Meyer-Mallat theorem associates an orthonormal wavelet basis to every orthonormal multiresolution analysis.

The Shannon scaling function is $\varphi (x)=\sin (\pi x)/(\pi x)$ and has a rectangular Fourier transform.

With the unnormalized Fourier transform, the scaling function with transform $1_{[-\pi ,\pi )}$ is $\varphi (x)=\sin (\pi x)/(\pi x)$. Its integer translates are orthonormal by orthonormal translates and Fourier periodization. Its refinement mask is the periodic indicator of $[-\pi /2,\pi /2)$, with $a_0=1$ and $a_n=2\sin (n\pi /2)/(\pi n)$. It belongs to $L^2$ but not $L^1$, so its transform is naturally interpreted through the Plancherel theorem.

A spatial Fourier mode has the form $e^{ikx}$ and is an eigenfunction of every constant-coefficient spatial differential operator; in particular, ${\partial }_x^2{e}^{ikx}=-k^2{e}^{ikx}$.

The Schwartz space $\mathcal{S}({\mathbb{R}}^n)$ consists of smooth functions whose derivatives decay faster than every inverse polynomial. Its Fréchet topology is generated by the seminormsIt is dense in every $H^s({\mathbb{R}}^n)$.

These seminorms measure all polynomially weighted derivatives and generate the Fréchet topology of the Schwartz space. Finitely many such bounds control each weighted derivative's integrable norm by adding a spatial decay factor stronger than $⟨x{⟩}^{-n}$. Continuity of the Fourier transform follows by applying those bounds after differentiation and integration by parts. Equivalent families use powers of $⟨x⟩$ instead of individual monomials.

For a non-Archimedean local field $F$, the Schwartz-Bruhat space $\mathcal{S}(F)$ consists of the locally constant, compactly supported complex-valued functions on $F$.

A Schwartz-Bruhat function on a non-Archimedean local field is a locally constant, compactly supported complex-valued function, or an element of the Schwartz-Bruhat space. A sufficiently small open additive subgroup leaves it invariant under translation, and only finitely many cosets meet its support.

A Schwartz function is an element of the Schwartz space: it is a smooth function and every derivative decreases faster than every inverse polynomial. Equivalently, $x^{\alpha }{D}^{\beta }f(x)$ defines a bounded function for every pair of multi-indices $\alpha ,\beta$.

A tempered distribution is a continuous linear functional on the Schwartz space. Tempered distributions include functions of at most polynomial growth and admit a Fourier transform by duality.

For $T\in {\mathcal{S}}^{\prime }$ and $\psi \in \mathcal{S}$, define $(T\ast \psi )(x)=⟨T_y,\psi (x-y)⟩$. Translations are smooth in the Schwartz space, so every derivative is ${\partial }^{\alpha }(T\ast \psi )(x)=⟨T,{\partial }^{\alpha }\psi (x-\cdot )⟩$. The finite-seminorm continuity estimate for $T$ implies $|{\partial }^{\alpha }(T\ast \psi )(x)|\le C_{\alpha }(1+|x|{)}^m$ for one $m$. Thus the result is a smooth function defining a tempered distribution, but it need not be a Schwartz function: $1\ast \psi =1$ when $\int \psi =1$. If both factors are radial, testing the rotation action on the translated $\psi$ shows that the convolution is radial.

For $n>1$, a tempered distribution $T$ on ${\mathbb{R}}^n$ is radial when $T\circ R=T$ for every $R\in SO(n)$, where $⟨T\circ R,\phi ⟩=⟨T,\phi \circ R^t⟩$. This agrees with the usual radial function definition for regular smooth distributions, because the special orthogonal group acts transitively on spheres.

Choose a radial mollifier $\rho \in C_c^{\infty }$ with integral one and support in the unit ball, and a radial cutoff function $\chi \in C_c^{\infty }$ equal to one there. For a radial tempered distribution $T$, the functionsare radial Schwartz functions. Smoothness comes from convolution of a tempered distribution with a Schwartz function, and compact support comes from the cutoff. Pairing with $\phi$ gives $⟨f_j,\phi ⟩=⟨T,{\check{\rho }}_{1/j}\ast (\chi (\cdot /j)\phi )⟩$. With $q_m(\phi )={\max }_{|\beta |\le m}{\sup }_x(1+|x|{)}^m|{\partial }^{\beta }\phi (x)|$, the cutoff tail and the mean value theorem giveThe continuity estimate for $T$ therefore proves $f_j\to T$ even in the strong dual topology. Inserting a cutoff is essential because convolution alone need not give rapid decay.

A sequence $u_j\in {\mathcal{S}}^{\prime }$ converges weakly to $u$ if $⟨u_j,\phi ⟩\to ⟨u,\phi ⟩$ for every Schwartz function $\phi$. This is the weak-star topology of the continuous dual of the Schwartz space, and differs from choosing a strong dual topology defined by uniform convergence on bounded sets of test functions.

With $\hat{\phi }(\xi )=\int e^{-ix\cdot \xi }\phi (x)dx$, the Fourier transform of a tempered distribution is defined byContinuity of the Fourier transform on the Schwartz space makes this well defined. For regular integrable functions, Fubini's theorem shows that it agrees with the ordinary transform. The inverse convention has factor $(2\pi {)}^{-n}$ and the positive exponential.

A real simple pole is not locally integrable in the ordinary improper sense. Its symmetric Cauchy principal value defines a tempered distribution. For the $e^{-ikx}$ transform convention, contour indentation or sine integration gives the displayed nondecaying oscillatory contribution. Upper and lower boundary-value prescriptions instead change it by $\mp i\pi \delta (x-a)$ before transforming. A pole prescription must therefore be stated before claiming a Fourier asymptotic expansion.

For an orthogonal matrix $R$, a unit-Jacobian change of variables gives $\widehat{\phi \circ R^t}=\hat{\phi }\circ R^t$. The dual definition of the Fourier transform of a tempered distribution consequently givesSince the Fourier transform is invertible on the Schwartz space and its dual, a tempered distribution is invariant under a rotation group exactly when its transform is invariant. This includes radial tempered distributions.

For $D_j=-i{\partial }_j$, the Fourier transform of a tempered distribution satisfiesThe base cases follow from $⟨D_{j}u,\phi ⟩=i⟨u,{\partial }_j\phi ⟩$ and integration by parts in the Fourier integral. Iterating gives the multi-index formulas. The $D$ convention matters: for the unscaled derivative, $\widehat{{\partial }_{j}u}=i{\xi }_j\hat{u}$.

For $u(x)=\frac{1}{2}\log (1+x^2)$, the angular-frequency transform is the tempered distributionThe numerator cancels the singularity at zero and makes this integral absolutely convergent. Differentiating $u$ and using $\widehat{(1+x^2{)}^{-1}}=\pi e^{-|\xi |}$ determines this expression up to a Dirac delta distribution. That delta coefficient is zero: testing with the expanding Gaussian ${\phi }_n(x)=(2\sqrt{\pi }{)}^{-1}{e}^{-x^2/(4n)}$ makes both the proposed expression and $⟨u,{\hat{\phi }}_n⟩$ tend to zero, while ${\phi }_n(0)$ stays fixed. Changing the subtraction convention would change the delta coefficient.

A real distribution $u$ is positive when $u[\varphi ]\ge 0$ for every nonnegative test function $\varphi$. Every positive distribution has order zero and is represented locally by a positive measure.

For the convention $\hat{f}(\xi )={\int }_{{\mathbb{R}}^n}{e}^{-ix\cdot \xi }f(x)dx$, if $f,\hat{f}\in L^1$, thenalmost everywhere. Reversing both exponential signs gives the equivalent opposite convention.

When $\hat{f}\in L^1$, its inverse Fourier integral is continuous by dominated convergence. If $f$ is also continuous and Fourier inversion identifies the two functions almost everywhere, then they agree everywhere.

For the angular-frequency convention and sufficient decay,

For the angular-frequency convention,Consequently cosine transforms to the half-sum of the two shifted deltas, while sine transforms to their signed difference divided by $2i$.

For the angular-frequency convention,so multiplication by a phase factor in frequency translates a function in physical space.

For ${\theta }_n(x)=(1-|x|/n{)}_{+}$,with value $n$ at zero.

The triangular functionhas angular-frequency Fourier transform $4{\sin }^2(\xi /2)/{\xi }^2$.

The squared sinc function is nonnegative and integrable; it occurs as the Fourier transform of a triangular function.

The cutoffs ${\theta }_n(\xi )=(1-|\xi |/n{)}_{+}$ increase pointwise to one, while their Fourier transforms are nonnegative and have integral $2\pi$ under the angular-frequency convention.

If $f\in L^1\cap L^{\infty }$ and $\hat{f}\ge 0$, testing against triangular frequency cutoffs and using monotone convergence gives $\Vert \hat{f}{\Vert }_1\le 2\pi \Vert f{\Vert }_{\infty }$.

The convolution of integrable functions on the real line is

A periodic convolution operator with continuous periodic kernel is compact on $L^2[0,2\pi ]$, since its kernel is square integrable. In the normalized Fourier series basis it is diagonal, with multipliers $c_n={\int }_0^{2\pi }K(x)e^{-inx}dx$. These multipliers include the full integral rather than its normalized Fourier coefficient.

When all multipliers are nonzero, the Moore–Penrose inverse of an operator acts by dividing each Fourier coefficient $y_n$ by $c_n$. Its data domain is precisely the square-summability condition displayed. The dense range need not be closed, and writing the formal inverse series does not establish the Picard criterion.

For a continuously differentiable periodic kernel, integration by parts and the Riemann-Lebesgue lemma give $c_n=o(1/|n|)$. Nonperiodic data $e^{\alpha x}$ with $\alpha \notin i\mathbb{Z}$ have Fourier coefficients asymptotic to a nonzero multiple of $1/n$. Their inverse coefficients $y_n/c_n$ therefore fail square summability, and there is no Hilbert-space inverse or least-squares minimizer. Periodic exponential data $\alpha =im$ instead give the single-mode solution $e^{imx}/c_m$.

If $v$ is a compactly supported distribution and $u$ is any distribution, define $⟨u\ast v,\varphi ⟩=⟨u_x,⟨v_y,\varphi (x+y)⟩⟩$. The inner function is smooth and supported in $\mathrm{supp}\varphi -\mathrm{supp}v$, hence is a test function for $u$. Finite-order estimates prove continuity, while the tensor product of distributions proves commutativity after inserting compact cutoffs. For every test function $\psi$, $(u\ast v)\ast \psi =u\ast (v\ast \psi )$; evaluating at zero against reflected test functions proves uniqueness. If both distributions have compact support, $\mathrm{supp}(u\ast v)\subset \mathrm{supp}u+\mathrm{supp}v$ and the convolution theorem gives $\widehat{u\ast v}=\hat{u}\hat{v}$. Without a compact factor or another support/growth condition, arbitrary distributional convolution is not generally defined.