Fourier analysis

Fourier analysis describes a function by the oscillations that compose it. A graph records values at successive positions or times; a frequency spectrum records how much of each frequency is present, together with its wave phase. The same complicated shape can become a short list of simple components. For periodic functions, the components form a Fourier series; on the whole real line, the Fourier transform uses a continuous range of frequencies, and the Fourier inversion theorem puts them back together.

For a concrete example, consider the trigonometric polynomialWith $t$ measured in seconds, its three Fourier harmonics have frequencies $1$, $3$, and $5$ cycles per second and amplitudes $1$, $1/2$, and $1/4$. Each cosine has zero wave phase. Adding them creates the sharper peaks in the figure. The frequency spectrum makes the recipe immediately visible; changing an amplitude changes the shape without changing the component's frequency. Fourier orthogonality lets us recover each Fourier coefficient by averaging against the corresponding sine or cosine, because the other Fourier harmonics cancel over a full period of a function.

The deeper reason these components are so useful is that they simplify operations. By Euler's formula, the exponential function $e^{i\omega x}$ packages a cosine and a sine into one Fourier mode. A translation or a derivative changes only its multiplier:Thus Fourier modes are simultaneous eigenfunctions of these operations. In Fourier series, this gives a literal diagonalization in the Fourier basis; on the real line, the Fourier transform gives the continuous counterpart, with each frequency handled separately. The Translation property of the Fourier transform and the Fourier transform of a derivative turn shifting and differentiating a whole function into multiplication of its Fourier transform. A convolution likewise becomes multiplication by the convolution theorem. This is beautiful because one description simplifies several apparently different operations at once.

For example, the heat equation $u_t=\kappa u_{xx}$, with thermal diffusivity $\kappa >0$, evolves a spatial Fourier mode asFine ripples have larger angular frequencies and decay faster: the formula explains why heating smooths a jagged temperature profile. For a general initial profile with sufficient regularity, the Fourier transform gives $\hat{u}(\omega ,t)=e^{-\kappa {\omega }^{2}t}\hat{u}(\omega ,0)$, and Fourier inversion recovers the temperature. Similarly, multiplying selected parts of a frequency spectrum suppresses unwanted components. Claude E. Shannon's Communication in the Presence of Noise (1949), Section III, describes ideal filtering as a projection that retains some frequency coordinates and removes others.^[ref]

The physical motivation is already visible in Joseph Fourier's Théorie analytique de la chaleur (1822). Chapter III constructs heat profiles using sums of sines and cosines; Chapter IX, articles 355–359, passes from Fourier series to integrals as the spatial interval becomes unbounded. These heat problems explain the motivation for the discrete and continuous descriptions. Their modern formulations specify which functions admit reconstruction and in what sense the expansions converge.^[ref][ref]

A Banach space $B\subseteq L^1(\mathbb{T})$ is homogeneous when its inclusion into $L^1$ is continuous, circle translations act isometrically, and each orbit is continuous in the $B$ norm. One commonly normalizes $\Vert f{\Vert }_1\le \Vert f{\Vert }_B$. The continuous functions with the supremum norm, and $L^p(\mathbb{T})$ for $1\le p<\infty$, are examples. Full $L^{\infty }$ fails translation continuity. If $g$ is continuous, the Bochner integral $f\ast g=\int g(t){\tau }_{t}fdt/(2\pi )$ belongs to $B$ and satisfies $\Vert f\ast g{\Vert }_B\le \Vert f{\Vert }_B\Vert g{\Vert }_1$. Approximation by continuous functions extends this to every $g\in L^1$.

The convolution estimate gives the upper bound. Apply the operator to the Fejér kernel, whose $L^1$ norm is one. The summability kernel on the circle property implies $f\ast K_n\to f$ in $B$, giving the reverse bound. This identifies the operator norm even when no single input attains it.

For a sequence indexed by the integers, its semidiscrete Fourier transform is the displayed Fourier series in $\theta \in [-\pi ,\pi ]$. For square-summable data the series is interpreted in the L2 norm. Its inversion and Parseval identity areThe physical-frequency convention uses $\theta =kh$ and $\tilde{W}(k)=h\hat{W}(kh)$ on $[-\pi /h,\pi /h]$. This is distinct from the discrete Fourier transform of a finite vector. Translation-invariant finite difference operators become Fourier multipliers, making Fourier stability analysis possible.

Consider an implicit centered heat stencil of width $\Delta x$ weighted by $\theta$, combined with an explicit centered stencil of width $2\Delta x$ weighted by $1-\theta$. With $\mu =\kappa \Delta t/(\Delta x{)}^2$, its amplification factor is $G(\xi )=[1-\mu (1-\theta ){\sin }^2\xi ]/[1+4\mu \theta {\sin }^2(\xi /2)]$. For $0\le \theta <1/2$, contraction in the discrete ${\ell }^2$ norm holds exactly under the displayed condition. It is unconditional for $1/2\le \theta \le 1$. Put $v={\sin }^2(\xi /2)$: the condition $G\ge -1$ reduces to $4\mu v[(1-2\theta )-(1-\theta )v]\le 2$. Maximizing this concave quadratic proves the threshold; the Parseval identity converts the multiplier bound into contraction.

A lower included wavenumber. A finite periodic patch of size $L$ has smallest nonzero wavenumber of order $2\pi /L$; supports and boundary conditions can change the factor. The cutoff limits large-scale fluctuations.

Fourier shell correlation compares two three-dimensional maps using their complex Fourier transform coefficients in successive reciprocal-space shells. Independent half-map refinement helps distinguish reproducible signal from overfitting; masking and correlated inputs can bias the result.

For this sign convention, cross-correlation compares a function with translated copies of another, conjugating the first factor. If $f,g\in L^1(\mathbb{R})$, Fubini's theorem gives Fourier transform $\widehat{f\star g}(k)=\overline{\hat{f}(k)}\hat{g}(k)$. Indeed, the substitution $s=x+y$ factors the double integral into $\int \overline{f(y)}{e}^{iky}dy$ and $\int g(s)e^{-iks}ds$. The zero-lag value for $f=g\in L^2$ is the squared L2 norm. Fourier inversion applied to this correlation proves the Parseval identity, first for Schwartz functions and then by density for L2 space.

A device or algorithm that estimates the cross-correlation of input signals, often by multiplying one signal by a delayed or complex-conjugated copy of another and averaging. In radio interferometry, calibrated complex correlations supply complex visibilities. Delay tracking, channel width and integration duration determine which field information survives that averaging.

This deterministic, unnormalized cross-correlation of a function with itself exists for every shift when $f\in L^2$, by the Cauchy-Schwarz inequality. It satisfies $h(0)=\Vert f{\Vert }_2^2$ and $h(-x)=\overline{h(x)}$. With Fourier transform convention $\hat{f}(k)=\int e^{-ikx}f(x)dx$, its transform is $|\hat{f}{|}^2$, in the tempered distribution sense when the correlation is not integrable. Fourier inversion at zero gives the Parseval identity, first for Schwartz functions and then by approximation in L2 space. Unlike the normalized statistical autocorrelation, its value at zero is not generally one.

The Wiener-Khinchin theorem relates the autocorrelation of a weakly stationary process to its power spectral density by a Fourier transform. The displayed convention is appropriate when the optical correlation is $\Gamma (\tau )=⟨f^{\ast }(t)f(t+\tau )⟩$ and the analytic field contains $e^{-i\omega t}$. A Lorentzian line shape of half-width $\kappa$ gives exponential correlation decay $e^{-\kappa |\tau |}$, including the carrier factor $e^{-i\omega \tau }$ when the full field is correlated.

A frequency spectrum describes the frequency components of a function. A Fourier series supplies discrete Fourier coefficients; the Fourier transform of an integrable function supplies a continuous, generally complex-valued description. An amplitude plot shows magnitudes and omits wave phases, so it need not determine the original function. In the example with three Fourier harmonics above, the one-sided amplitude plot has heights $1$, $1/2$, and $1/4$. The corresponding complex Fourier coefficients at positive and negative frequencies each have half those heights: $\cos (2\pi \nu t)=(e^{2\pi i\nu t}+e^{-2\pi i\nu t})/2$.

A spectrogram displays how a local frequency spectrum changes over time, usually using colour for squared Fourier transform magnitude. Longer analysis windows give finer frequency spacing while averaging over a longer time interval.

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.

The Nyquist–Shannon sampling theorem says that a bandlimited function can be recovered exactly from equally spaced samples, provided the sampling rate is high enough. It is the bridge between a continuous waveform and a sequence of numbers: the missing values between samples are determined by the restriction on its frequencies.

Here is a precise finite-energy version. Let $f\in L^2(\mathbb{R})$, meaning ${\int }_{\mathbb{R}}|f(t){|}^{2}dt<\infty$, and suppose its Fourier transform, defined through the Plancherel theorem, is zero outside $[-2\pi B,2\pi B]$. Thus $B>0$ is the largest allowed ordinary frequency, in cycles per second, rather than an angular frequency. Set $T=1/f_s$, where $f_s$ is the number of samples per second. If $f_s\ge 2B$, the continuous representative of $f$ is recovered byThe quotient is the normalized sinc function, with value one when its argument is zero. Each shifted sinc function is one at its own sampling time and zero at every other sampling time. Its coefficient is therefore exactly the required sample value. For this L2 space formulation, the reconstruction converges in the L2 space norm and uniformly in $t$. The critical rate $f_s=2B$ is allowed for this class; choosing $f_s>2B$ leaves a gap between the occupied frequencies and the sampling boundary.

For a numerical example, a bandlimited function with no frequencies above $3$ cycles per second can be reconstructed from its samples every $0.1$ seconds: $f_s=10>6$. Exact reconstruction uses the whole sequence indexed by $\mathbb{Z}$. A finite recording and a finite sum of sinc functions introduce truncation error; joining adjacent samples by straight lines is generally a different reconstruction.

The obstruction is aliasing. At sampling rate $f_s=10$, the two distinct waveformshave identical samples at $t=n/10$, becauseThe samples alone cannot tell whether the waveform oscillates twice or eight times per second. If we know in advance that all frequencies are below $5$, the $8$-cycle waveform is excluded and the ambiguity disappears. These pure cosines are an illustrative aliasing example, rather than members of the finite-energy L2 space class in the stated theorem.

Exactly at a sampling boundary, the function class matters. At rate $f_s$, the sine $\sin (2\pi (f_s/2)t)$ has zero values at every $n/f_s$. It is a nonzero waveform indistinguishable from zero on that grid, but has infinite energy and is excluded by the finite-energy hypothesis above. This explains why a rule intended to cover arbitrary pure sinusoids should demand a strict inequality. For spatial data, replace seconds by distance: the Nyquist spatial frequency is $1/(2p)$ for sample pitch $p$.

The proof idea is short: the samples are Fourier coefficients of the Fourier transform restricted to a sampling-sized frequency interval. Because a bandlimited function has no contribution outside that interval, its Fourier series determines the whole Fourier transform; Fourier inversion then gives the sinc function expansion. See sampling expansion by periodic Fourier projection for this connection.

Claude E. Shannon's Communication in the Presence of Noise (1949), Section II, Theorem 1 and equation (7), gives the sampling statement and reconstruction using shifted sinc functions in a communication setting. Shannon explicitly notes earlier mathematical forms and credits Nyquist with emphasizing the sampling interval in telegraphy.^[ref]

For uniform samples at rate $f_s$, the frequency $f_s/2$ is the Nyquist frequency. A real baseband signal confined below this boundary can avoid spectral overlap when an appropriate anti-aliasing filter is used.

Aliasing is the inability to distinguish different continuous frequencies from sampled values. The identity above says that frequencies separated by an integer multiple of the sampling rate $f_s$ give identical samples of a complex Fourier mode. For real cosines, changing the sign of a frequency also gives the same waveform, explaining the $2$-versus-$8$ example at rate $10$. The Nyquist–Shannon sampling theorem prevents this ambiguity within its specified class of bandlimited functions. Before sampling, unwanted higher frequencies must be suppressed if they would violate that assumption; once two waveforms give the same samples, those samples cannot identify which was present.

Ambiguity between spatial oscillations sampled on a grid. At positions $x_j=j\Delta x$, the factors $e^{ik{x}_j}$ and $e^{i(k+2\pi m/\Delta x)x_j}$ agree for every integer $m$. Spatial wavenumbers outside $|k|<\pi /\Delta x$ therefore cannot be distinguished without additional assumptions. For a traveling event of temporal frequency $f$ and apparent speed $v_{\mathrm{app}}$, the associated criterion is $\Delta x<v_{\mathrm{app}}/(2f)$, with a margin for filtering. Temporal oversampling does not remove this spatial ambiguity.

An anti-aliasing filter suppresses signal components outside the intended acquisition band before sampling. For voltage recording it is an analogue filter ahead of the analog-to-digital converter, with a transition band allowed below half the sampling rate. Once two frequencies have become indistinguishable through aliasing, a digital filter generally cannot recover the lost distinction. The Nyquist–Shannon sampling theorem concerns an already bandlimited function; a physical filter approximates that prerequisite.

For every $ϵ>0$, choose a nonzero smooth spectral bump supported in $(-{ϵ}_0,{ϵ}_0)$ with ${ϵ}_0<\min (ϵ,\pi )$, and let $h$ be its inverse Fourier transform. Then $g$ is a nonzero Schwartz function, vanishes on all integer samples, and its transform is the difference of bumps shifted to $\pi$ and $-\pi$. Its support lies strictly within $[-\pi -ϵ,\pi +ϵ]$. Thus even a small bandwidth increase destroys uniqueness from unit-spaced samples.

When the continuous Fourier transform $F$ of an integrable function vanishes outside $[-\pi ,\pi ]$, its periodic Fourier coefficients are the integer samples of the continuous representative 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 reconstruction using the normalized sinc function, with each quotient defined as one when $t=n$. The Cauchy-Schwarz inequality controls the reconstruction error uniformly in $t$. Bessel's inequality bounds the vector of shifted sinc functions, 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.

A sublattice is a subgroup of a Euclidean lattice. It is a Euclidean lattice in its real linear span, which may have smaller dimension than the original ambient Euclidean space. The equal-parity integer points in ${\mathbb{Z}}^d$ form a full-rank sublattice of index $2^{d-1}$: reduction of all coordinate differences modulo two gives the quotient.

For an integer basis of a Euclidean lattice, the algorithm uses integer column operations and swaps to obtain a size-reduced basis with Gram-Schmidt coefficients $|{\mu }_{ij}|\le 1/2$ and the condition $\delta \Vert b_i^{\ast }{\Vert }^2\le \Vert b_{i+1}^{\ast }{\Vert }^2+{\mu }_{i+1,i}^2\Vert b_i^{\ast }{\Vert }^2$. It preserves the lattice and exposes short vectors and near-relations. It is useful for improving proven Diophantine bounds by encoding scaled logarithms in integer coordinates. A certified lower bound for every nonzero lattice vector is ${\min }_i\Vert b_i^{\ast }\Vert$: in a nonzero integer linear combination, project onto the last nonzero Gram-Schmidt direction. Better-reduced bases improve this lower bound. Rounding and logarithm errors must still be bounded before turning it into a certified lower bound for a logarithmic form.

A complex lattice is a discrete rank-two subgroup of the complex plane, generated by two real-linearly independent periods. An oriented basis can be chosen with $\mathrm{Im}({\omega }_1/{\omega }_2)>0$. Scaling by ${\omega }_2^{-1}$ then gives $\mathbb{Z}\tau +\mathbb{Z}$ for a point in the complex upper half-plane.

Two complex lattices are homothetic when multiplication by one nonzero complex number takes one onto the other. Changes between oriented integral bases have determinant one, so normalized period ratios classify these classes by $S{L}_2(\mathbb{Z})\backslash \mathbb{H}$. The lattice Eisenstein sums scale as $G_k(a\Lambda )=a^{-k}{G}_k(\Lambda )$.

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.

Choose positive axis weights linearly independent over the rationals, and suppose both integer Euclidean lattices contain $q{\mathbb{Z}}^d$ for some positive integer $q$. A length-preserving linear map sends each $q{e}_i$ to an integer vector with exactly that absolute-coordinate pattern, by the displayed implication. Thus the map must be a diagonal sign change. For the four-coordinate tetracode, preserving the code forces all four signs to agree, because each of its four nonzero projective words has support on a different triple of coordinates. A global sign cannot exchange the two odd-vector parity conditions, proving nonisometry for these weights.

Add respectively the conditions ${\sum }_i{x}_i\equiv 0(\mathrm{mod}4)$ and ${\sum }_i{x}_i-2x_1\equiv 0(\mathrm{mod}4)$ to the displayed common conditions. These define the two integer Euclidean lattices underlying the tetracode length-isospectral lattice construction. They coincide on their all-even vectors and differ on all-odd vectors. Reflecting a coordinate divisible by three preserves the code condition; on odd vectors it changes the coordinate sum by two modulo four, while on even vectors both conditions remain satisfied. This yields the coordinate reflection bijection for tetracode lattices.

For the tetracode length-isospectral lattice construction $L_{\pm }=(\pm 3I+S){\mathbb{Z}}^4$, choose the first coordinate $i$ divisible by three and reverse its sign. Such a coordinate always exists because each nonzero tetracode word has one zero coordinate. The reflection preserves the reduction modulo three and switches the two lattice membership conditions modulo four. It is an involution between $L_{+}$ and $L_{-}$ preserving every absolute coordinate, so arbitrary positive diagonal weights preserve equality of their vector-length multisets.

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 of a full-rank Euclidean lattice $\Lambda \subset {\mathbb{R}}^n$ isHere $⟨x,y⟩$ is the Euclidean inner product. If the columns of an invertible matrix $B$ form a basis of $\Lambda =B{\mathbb{Z}}^n$, thenIndeed, $y$ belongs to the dual lattice exactly when $B^{\mathsf{T}}y\in {\mathbb{Z}}^n$; $B^{-\mathsf{T}}$ denotes the transpose of the matrix inverse. The reciprocal covolume follows by taking the determinant. With the Fourier transform kernel $e^{-2\pi i⟨x,y⟩}$, these are precisely the frequencies of the characters of a real torus: translating $x$ by any $\lambda \in \Lambda$ leaves $e^{2\pi i⟨x,y⟩}$ unchanged. In one dimension, $(h\mathbb{Z}{)}^{\vee }=h^{-1}\mathbb{Z}$ for $h>0$: increasing the spacing in physical space decreases the spacing in frequency space.

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.

The Poisson summation formula equates sample sums of a function and its Fourier transform. Exchanging broad and narrow sums can save work. Use the conventionFor every Schwartz function $f\in \mathcal{S}(\mathbb{R})$,Both series have absolute convergence. A sufficient hypothesis is that $f$ is smooth and ${\sup }_x(1+|x|{)}^M|f^{(r)}(x)|<\infty$ for all nonnegative integers $M,r$; its Fourier transform is then also a Schwartz function.^[ref] This assumption controls both decay and regularity; mere membership of an L1 space does not control values at individual sampling points, as periodization of an integrable function illustrates.

Here $\xi$ counts cycles per unit length. In the angular convention used elsewhere in Fourier analysis, $\tilde{f}(\omega )=\int f(x)e^{-i\omega x}dx$, so $\hat{f}(\xi )=\tilde{f}(2\pi \xi )$ and the right-hand side becomes ${\sum }_k\tilde{f}(2\pi k)$. Keeping the Fourier transform convention explicit prevents stray $2\pi$ factors. Changing the sample spacing leads to the Poisson summation formula for a Euclidean lattice.

Take the Gaussian function $f_a(x)=e^{-\pi a{x}^2}$ with $a>0$. Every derivative is a polynomial times the same Gaussian function, so $f_a$ is a Schwartz function. The Fourier transform of a Gaussian gives, in the cycles convention,Thus the Poisson summation formula gives the exact identityThe Gaussian Fourier transform is given in scaled form in Exercise 41 of Terence Tao's 245C, Notes 2: The Fourier transform.^[ref] The identity is also the imaginary-axis transformation of the Jacobi theta function. At $a=1$ the Gaussian function is its own Fourier transform; at reciprocal values of $a$ its spatial and spectral widths exchange.

For a short derivation of the Gaussian Fourier transform, write $H(\xi )={\hat{f}}_a(\xi )$. Differentiation under the integral and integration by parts, using $f_a^{\prime }(x)=-2\pi ax{f}_a(x)$, giveThe Gaussian integral gives $H(0)=a^{-1/2}$, so this ordinary differential equation yields the claimed formula. The decay of the Gaussian function justifies these steps and removes the boundary term.^[ref]

For $a=0.04$, the physical Gaussian function has width parameter $1/\sqrt{a}=5$, while its Fourier transform has width parameter $\sqrt{a}=0.2$. Many physical samples matter, but the dual zero sample already givesFor $a=25$, use the physical side instead:These are bounds on the full positive tails, obtained below, rather than estimates based only on the first omitted term. Ordinary floating-point arithmetic rounds these sums to $5$ and $1$; the exact real numbers are slightly larger. A Gaussian sum truncation bound lets us choose the cheaper side and certify its truncation error before evaluating it.

For $b>0$ and a nonnegative integer $N$, define the symmetric partial sum and its positive remainder bySuccessive omitted terms have ratio $e^{-\pi b(2n+1)}\le e^{-\pi b(2N+3)}$ for $n\ge N+1$. Bounding them by a geometric series givesConsequently the physical approximation to $S(a)$ has error at most $B_N(a)$, while the dual approximation $a^{-1/2}{S}_N(1/a)$ has error at most $a^{-1/2}{B}_N(1/a)$. The prefactor must be included when setting an absolute error tolerance.

For a truncation error below $10^{-12}$, at $a=0.04$ the physical side first passes this bound with $N=15$: $31$ lattice points give $B_{15}(0.04)<2.172\times 10^{-14}$, whereas $N=14$ gives $B_{14}(0.04)>10^{-12}$. The dual side needs only $k=0$. At $a=25$ the physical side needs only $n=0$, whereas the dual side first passes the bound with $N=14$: $29$ lattice points give $25^{-1/2}{B}_{14}(0.04)<2.146\times 10^{-13}$. These counts measure symmetric lattice terms; using the even function symmetry reduces the distinct exponential function evaluations to $N+1$. Truncation bounds do not include rounding error in a numerical implementation.

For $F\in L^1(\mathbb{R})$ the periodization is absolutely convergent almost everywhere in a period by integrating the sum of absolute values. Its Fourier series coefficient is $c_n=\tilde{F}(n)/(2\pi )$, where $\tilde{F}(\omega )={\int }_{\mathbb{R}}F(x)e^{-i\omega x}dx$. Absolute integrability permits integrating term by term and changing variables across the disjoint period intervals. This coefficient identity does not by itself assert pointwise convergence of the ordinary Fourier series. Even uniform convergence of the original periodization cannot repair arbitrary point values: $F=1_{\{0\}}$ has zero transform and a nonzero periodization at multiples of $2\pi$. Additional regularity, such as a continuous piecewise continuously differentiable periodization, validates the pointwise Poisson summation formula in this setting.

// Target: fluid-mechanics.bigb

For a Schwartz function on the real numbers, its periodization $P_f$ is a smooth periodic function of period of a function one. Rapid decay gives uniform convergence on $[0,1]$ of the defining series and each differentiated series. Its $k$th Fourier coefficient isThe change of variable $u=x+n$ uses $e^{2\pi ikn}=1$. Rapidly decreasing Fourier coefficients give the absolutely and uniformly convergent Fourier series $P_f(x)={\sum }_k\hat{f}(k)e^{2\pi ikx}$. Evaluation at $x=0$ proves the Poisson summation formula.^[ref]

The same identity explains aliasing in the Nyquist–Shannon sampling theorem: applying the Poisson summation formula to the Schwartz function $x\mapsto f(x)e^{-2\pi i\xi x}$ givesThe sampled spectrum is periodic in frequency, and shifted copies of the continuous Fourier transform add together. When those copies overlap, the samples cannot separate their contributions. The Nyquist–Shannon sampling theorem adds a bandlimited function hypothesis to make reconstruction possible.

Let $\Lambda =B{\mathbb{Z}}^d$ be a full-rank Euclidean lattice, with covolume $m(\Lambda )=|\det B|$ and dual lattice ${\Lambda }^{\vee }=B^{-\mathsf{T}}{\mathbb{Z}}^d$. For a Schwartz function $f\in \mathcal{S}({\mathbb{R}}^d)$ use the Fourier transform conventionThen the Poisson summation formula becomesBoth series have absolute convergence. To see the normalization, apply integer-lattice Poisson summation to $g(u)=f(Bu)$. The change of variables formula gives $\hat{g}(k)=|\det B{|}^{-1}\hat{f}(B^{-\mathsf{T}}k)$, yielding both the reciprocal covolume factor and the dual lattice argument. The factor is the physical lattice-point density. In particular, for spacing $h>0$,For the Gaussian function $f(x)=e^{-\pi a\Vert x{\Vert }^2}$ with $a>0$, the product of the one-dimensional Gaussian Fourier transforms givesHere $\Vert \cdot \Vert$ is the Euclidean norm. As a normalization check, take $\Lambda =\mathrm{diag}(2,3){\mathbb{Z}}^2$. Its covolume is $6$, and its dual lattice is $\mathrm{diag}(1/2,1/3){\mathbb{Z}}^2$, soThis also follows by multiplying two Gaussian Poisson summation identities with parameters $4a$ and $9a$. If one instead uses an angular Fourier transform kernel $e^{-i⟨x,\omega ⟩}$, the reciprocal frequencies are $2\pi {\Lambda }^{\vee }$, while the prefactor remains $m(\Lambda {)}^{-1}$.

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 $j=0$.