Poisson Resonance Formula

• The Poisson resonance formula is a generalization of classical Poisson summation, replacing the Fourier transform with unitary operators defined by arithmetic weight sequences.
• It uses weighted dilation operators and Dirichlet convolution inverses to establish a pointwise operator identity that reflects spectral symmetry.
• The construction reveals differential covariance and a spectral correspondence via Mellin diagonalization, extending Fourier transform properties in an arithmetic framework.

The Poisson resonance formula is a weighted generalization of the classical Poisson summation formula in which the Fourier transform is replaced by a family of unitary operators determined by an arithmetic weight sequence $(a_n)$. In the formulation developed in "A Family of Unitary Operators Satisfying a Poisson-type Summation Formula" (Faifman, 2011), the central object is a Fourier–Poisson operator $F(a_n)$ acting on $L^2[0,\infty)$ and satisfying the operator identity

$T(a_n)F(a_n)=S\,T(a_n),$

where

$$T(a_n)f(x)=\sum_{n\ge 1} a_n f(nx), \qquad Sf(x)=x^{-1}f(1/x).$$

Under suitable hypotheses, this yields the pointwise formula

$$\sum_{n\ge1} a_n\,F(a_n)f(nx)=\frac1x\sum_{n\ge1} a_n\,f(n/x),$$

which is the paper’s Poisson-type summation identity in operator form. Conceptually, the formula exhibits a family of unitary involutions whose spectral data are encoded by a Dirichlet series on the critical line.

1. Classical antecedent and general framework

The classical Poisson summation formula enters the theory through even functions, equivalently through the model $L^2[0,\infty)$. The restriction to even functions is natural because the Fourier transform preserves parity and the Poisson formula is trivial on odd functions. In the classical setting, for even $f$ with $f(0)=0$,

$$\sum_{n\ge1} f(nx)=\frac1x\sum_{n\ge1}\widehat f(n/x).$$

The generalized theory replaces the unweighted dilation sum by a weighted operator $F(a_n)$0, and replaces the Fourier transform by an operator $F(a_n)$1 adapted to the coefficients $F(a_n)$2 (Faifman, 2011).

The basic arithmetic input is the existence of a Dirichlet convolution inverse $F(a_n)$3 for $F(a_n)$4. This means

$F(a_n)$5

Equivalently, if

$F(a_n)$6

then

$F(a_n)$7

This converts the weighted dilation-sum operator into the analogue of an invertible Fourier-side object. The generalized Poisson formula is therefore not merely a deformation of the classical identity; it is an operator-theoretic construction tied to Dirichlet series and arithmetic convolution.

2. Weighted dilation operators and the Poisson identity

The operator

$F(a_n)$8

is initially formal, and its analytic meaning depends on decay of the coefficients. The key hypothesis is

$F(a_n)$9

together with the analogous condition

$L^2[0,\infty)$0

for the inverse coefficients. Under these assumptions, $L^2[0,\infty)$1 extends to a bounded operator on $L^2[0,\infty)$2, and Lemma 3.1 gives the norm estimate

$L^2[0,\infty)$3

The same applies to $L^2[0,\infty)$4 (Faifman, 2011).

With these operators available, the Fourier–Poisson operator is defined by

$L^2[0,\infty)$5

Formally this is $L^2[0,\infty)$6, so the inversion $L^2[0,\infty)$7 is conjugated by the weighted dilation structure. The resulting operator satisfies

$L^2[0,\infty)$8

which is the generalized Poisson identity at the operator level. For sufficiently regular $L^2[0,\infty)$9, this yields the pointwise summation formula

$T(a_n)F(a_n)=S\,T(a_n),$0

This formula is the precise analogue of the classical Poisson symmetry, but now indexed by an arbitrary arithmetic weight system possessing a Dirichlet inverse.

3. Unitarity and the family of Fourier–Poisson transforms

A central result is Theorem 3.3, which proves that under the strong decay assumptions on $T(a_n)F(a_n)=S\,T(a_n),$1 and $T(a_n)F(a_n)=S\,T(a_n),$2, the operator $T(a_n)F(a_n)=S\,T(a_n),$3 is unitary on $T(a_n)F(a_n)=S\,T(a_n),$4 (Faifman, 2011). At the formal level, $T(a_n)F(a_n)=S\,T(a_n),$5 is an involution because $T(a_n)F(a_n)=S\,T(a_n),$6 on a dense test space, and the theorem shows that this structure persists in the Hilbert-space setting.

The construction therefore produces not a single generalized transform, but a whole family of unitary operators parameterized by arithmetic data. The role of the Fourier transform in classical Poisson summation is taken by $T(a_n)F(a_n)=S\,T(a_n),$7, while the role of the lattice-sum symmetry is taken by

$T(a_n)F(a_n)=S\,T(a_n),$8

The classical Fourier transform appears as a special formal case, whereas other choices of coefficients give different unitary operators. The paper specifically notes examples with alternating dyadic coefficients.

This family viewpoint is important. The generalized formula does not merely preserve the visible summation pattern; it preserves the deeper unitary symmetry behind Poisson summation. This suggests that the resonance terminology refers to a structural matching between weighted dilation sums and the inversion operator $T(a_n)F(a_n)=S\,T(a_n),$9, mediated by a unitary transform whose spectral behavior is governed by a Dirichlet series.

4. Mellin diagonalization, weak existence, and uniqueness

A major structural result is the diagonal form of the weighted operators in Mellin/Fourier coordinates. The paper introduces

$$T(a_n)f(x)=\sum_{n\ge 1} a_n f(nx), \qquad Sf(x)=x^{-1}f(1/x).$$0

then

$$T(a_n)f(x)=\sum_{n\ge 1} a_n f(nx), \qquad Sf(x)=x^{-1}f(1/x).$$1

and sets $$T(a_n)f(x)=\sum_{n\ge 1} a_n f(nx), \qquad Sf(x)=x^{-1}f(1/x).$$2. Conjugation by $$T(a_n)f(x)=\sum_{n\ge 1} a_n f(nx), \qquad Sf(x)=x^{-1}f(1/x).$$3 diagonalizes the dilation-sum operator: $$T(a_n)f(x)=\sum_{n\ge 1} a_n f(nx), \qquad Sf(x)=x^{-1}f(1/x).$$4 Thus $$T(a_n)f(x)=\sum_{n\ge 1} a_n f(nx), \qquad Sf(x)=x^{-1}f(1/x).$$5 becomes multiplication by the Dirichlet series evaluated on the critical line (Faifman, 2011).

From this diagonalization it follows that

$$T(a_n)f(x)=\sum_{n\ge 1} a_n f(nx), \qquad Sf(x)=x^{-1}f(1/x).$$6

almost everywhere. The paper identifies this with an arithmetic orthogonality condition on the coefficients, given explicitly in Proposition 4.1.

The paper also proves a weaker existence theorem. Under the milder hypothesis

$$T(a_n)f(x)=\sum_{n\ge 1} a_n f(nx), \qquad Sf(x)=x^{-1}f(1/x).$$7

there still exists a bounded unitary operator $$T(a_n)f(x)=\sum_{n\ge 1} a_n f(nx), \qquad Sf(x)=x^{-1}f(1/x).$$8 satisfying the operator-form identity

$$T(a_n)f(x)=\sum_{n\ge 1} a_n f(nx), \qquad Sf(x)=x^{-1}f(1/x).$$9

even when the stronger pointwise Poisson formula is no longer available for all sequences. In diagonal coordinates the operator is

$$\sum_{n\ge1} a_n\,F(a_n)f(nx)=\frac1x\sum_{n\ge1} a_n\,f(n/x),$$0

or, wherever the quotient has modulus one,

$$\sum_{n\ge1} a_n\,F(a_n)f(nx)=\frac1x\sum_{n\ge1} a_n\,f(n/x),$$1

with phase $$\sum_{n\ge1} a_n\,F(a_n)f(nx)=\frac1x\sum_{n\ge1} a_n\,f(n/x),$$2 at zeros of the denominator. Theorem 5.2 states that this operator is unitary, and if

$$\sum_{n\ge1} a_n\,F(a_n)f(nx)=\frac1x\sum_{n\ge1} a_n\,f(n/x),$$3

for some $$\sum_{n\ge1} a_n\,F(a_n)f(nx)=\frac1x\sum_{n\ge1} a_n\,f(n/x),$$4, then such a bounded operator satisfying $$\sum_{n\ge1} a_n\,F(a_n)f(nx)=\frac1x\sum_{n\ge1} a_n\,f(n/x),$$5 is unique.

The distinction between the strong and weak regimes is conceptually significant. In the strong regime one has a pointwise summation identity; in the weak regime one still has the exact operator-theoretic analogue of Poisson summation. This suggests that the operator identity, rather than the termwise formula, is the fundamental object.

5. Interaction with differentiation

The generalized Fourier–Poisson operators reproduce a key feature of the ordinary Fourier transform: covariance with differentiation. The paper introduces the symmetric differential operator

$$\sum_{n\ge1} a_n\,F(a_n)f(nx)=\frac1x\sum_{n\ge1} a_n\,f(n/x),$$6

with

$$\sum_{n\ge1} a_n\,F(a_n)f(nx)=\frac1x\sum_{n\ge1} a_n\,f(n/x),$$7

For Schwartz-type functions in a suitable class $$\sum_{n\ge1} a_n\,F(a_n)f(nx)=\frac1x\sum_{n\ge1} a_n\,f(n/x),$$8, the operator $$\sum_{n\ge1} a_n\,F(a_n)f(nx)=\frac1x\sum_{n\ge1} a_n\,f(n/x),$$9 satisfies

$L^2[0,\infty)$0

(Faifman, 2011).

This is the generalized analogue of the classical Fourier identity $L^2[0,\infty)$1. The proof uses analytic continuation in a strip and a growth condition on the Dirichlet-series ratio

$L^2[0,\infty)$2

which must satisfy polynomial bounds in $L^2[0,\infty)$3 uniformly for $L^2[0,\infty)$4 in a strip around $L^2[0,\infty)$5. Under these assumptions, $L^2[0,\infty)$6 preserves the appropriate Schwartz class and anticommutes with $L^2[0,\infty)$7.

The differentiation result clarifies that the construction is not limited to a summation identity. It extends part of the full operational calculus associated with the Fourier transform. In this sense the Poisson resonance formula sits inside a broader harmonic-analytic framework in which arithmetic Dirichlet data dictate both unitary symmetry and differential covariance.

6. Terminological scope and related formulas

In the usage of (Faifman, 2011), the Poisson resonance formula is the weighted operator identity

$L^2[0,\infty)$8

and its pointwise realization through weighted dilation sums. The term is not a universally fixed designation across mathematics. Other literatures use closely related Poisson language for different resonance-type identities.

In scattering theory, Poisson formulas express wave traces in terms of resonances. For perturbed three-dimensional Dirac operators, the wave trace

$L^2[0,\infty)$9

is represented as a distributional sum over resonances and bound states away from $f$0 (Kungsman et al., 2014). For Schwarzschild–de Sitter backgrounds, an exact trace expansion relates the wave propagator to scattering resonances together with explicit correction terms coming from non-compact exponential tails (Oltman et al., 18 Dec 2025). In a different direction, the Poisson–Newton formula associates to a meromorphic Dirichlet series a distributional identity between exponentials indexed by zeros and poles and atomic masses indexed by the Dirichlet frequencies (Muñoz et al., 2013).

This suggests that the expression Poisson resonance formula is context-dependent. In the weighted Fourier–Poisson theory of (Faifman, 2011), it denotes a nonclassical summation symmetry implemented by a unitary operator whose spectral multiplier is $f$1. In resonance theory and trace-formula settings, analogous language refers instead to exact expansions of wave or spectral data in terms of resonance sets. The common feature is the replacement of classical Fourier duality by a more structured spectral correspondence.