Angular Harmonic Space Analysis: PTA & Cosmology

• Angular harmonic space analysis is a method that transforms real-space observables into multipole or angular-momentum channels, isolating distinct modes and clarifying covariance structure.
• It underpins precise analyses in pulsar timing arrays, enabling the extraction of quadrupolar gravitational-wave signatures and refined factorization of the Earth-term response.
• The approach extends to cosmology and nonlinear optics, facilitating power-spectrum estimation, mode decomposition, and the study of orbital angular momentum in structured light.

Searching arXiv for recent and foundational papers on angular harmonic-space analysis across PTA, cosmology, and structured-light contexts. arXiv_search(query="angular harmonic space analysis spherical harmonics harmonic space pulsar timing array structured light", max_results=10, sort_by="relevance") to=arxiv.search code 򐂕json {"query":"angular harmonic space analysis spherical harmonics harmonic space pulsar timing array structured light", "max_results": 10, "sort_by": "relevance"} Searching more specifically for PTA harmonic-space treatments and harmonic analysis in structured-light harmonic generation. arXiv_search(query="pulsar timing array harmonic analysis angular correlations harmonic space structured light high harmonic generation orbital angular momentum", max_results=12, sort_by="relevance") to=arxiv.search code .awtextrajson {"query":"pulsar timing array harmonic analysis angular correlations harmonic space structured light high harmonic generation orbital angular momentum", "max_results": 12, "sort_by": "relevance"} Angular harmonic space analysis denotes a class of methods in which angular dependence is represented in basis functions adapted to symmetry, and then analyzed through the corresponding coefficients rather than only through real-space observables. Across the cited literature, this includes expansions in Legendre polynomials for pair-separation correlations, spherical and spin-weighted spherical harmonics on the sphere, Cartesian harmonic tensors as completely symmetric traceless tensors, and vortex factors such as $e^{i\ell\phi}$ or $e^{i\ell\phi_{st}}$ for spatial and spatiotemporal phase structure. This suggests a common methodological core: angular information is reorganized into multipoles or angular-momentum channels, which often diagonalizes the response, isolates physically distinct modes, and makes covariance structure explicit (Agazie et al., 2024, Roebber et al., 2016, Allen, 2024, Parke, 2023, Gui et al., 2021).

1. Basis functions, multipoles, and diagonalization

In pulsar timing array (PTA) analyses, the angular correlation function can be expanded in Legendre polynomials,

$$C(\theta)=\sum_{\ell=0}^\infty C_\ell \frac{2\ell+1}{4\pi}P_\ell(\cos\theta),$$

so that the real-space two-point function is the harmonic transform of the angular power spectrum (Roebber et al., 2016). In the NANOGrav harmonic analysis, the Hellings–Downs correlation is written as

$$\Gamma^{\rm HD}_{ab}=(1+\delta_{ab})\sum_{\ell=2}^{\infty} c_\ell^{\rm HD} P_\ell(\cos\theta_{ab}),$$

with

$$c_\ell^{\rm HD}=\frac{3}{2}(2\ell+1)\frac{(\ell-2)!}{(\ell+2)!},\qquad \ell\ge 2,$$

and $c_0^{\rm HD}=c_1^{\rm HD}=0$. The first nonzero multipole is therefore the quadrupole, and the coefficient drops rapidly at higher $\ell$ roughly as $\ell^{-3}$ (Agazie et al., 2024).

For the PTA response itself, harmonic space provides a more refined factorization. The Earth-term redshift response can be written in a “diagonalized form”,

$$F(\Omega;\Omega_p)=\sum_{l=2}^{\infty}\sum_{m=-l}^{l} A_l\,{}_2Y_{lm}(\Omega)\,Y_{lm}(\Omega_p), \qquad A_l=\frac{1}{\sqrt{(l+2)(l+1)l(l-1)}},$$

so that the gravitational-wave direction is carried by spin-2 harmonics and the pulsar direction by ordinary spin-0 harmonics (Allen, 2024). The same $l,m$ labels appear on both sides, which makes orthogonality manipulations straightforward and exposes the spin-2 structure of the response.

A Cartesian reformulation exists when angular dependence is more naturally handled tensorially. Cartesian harmonic tensors are “completely symmetric traceless tensors in three dimensional space constructed from the direct product of a unit vector with itself.” In $e^{i\ell\phi_{st}}$0, the number of independent components is $e^{i\ell\phi_{st}}$1, matching the multiplicity of $e^{i\ell\phi_{st}}$2, and angular couplings can then be written algebraically in terms of scalar products of unit vectors (Parke, 2023). This provides a basis-equivalent alternative to spherical harmonics when tensor contractions or multi-harmonic couplings are the primary objects.

2. Pulsar timing arrays and the angular structure of a gravitational-wave background

PTA work provides one of the clearest demonstrations of angular harmonic space analysis as an inference tool. The stochastic background model writes the cross-power spectral density as

$e^{i\ell\phi_{st}}$3

with angular dependence isolated in $e^{i\ell\phi_{st}}$4 (Agazie et al., 2024). General relativity predicts the Hellings–Downs shape, whose harmonic content begins at $e^{i\ell\phi_{st}}$5, so the quadrupole is the leading angular signature of an isotropic stochastic gravitational-wave background in GR (Agazie et al., 2024).

A harmonic-space reconstruction with free Legendre coefficients $e^{i\ell\phi_{st}}$6 showed that, for the NANOGrav 15-year data set, the quadrupole-only model $e^{i\ell\phi_{st}}$7 has Bayes factor $e^{i\ell\phi_{st}}$8 relative to common uncorrelated red noise, and the inferred coefficient satisfies

$e^{i\ell\phi_{st}}$9

consistent with $$C(\theta)=\sum_{\ell=0}^\infty C_\ell \frac{2\ell+1}{4\pi}P_\ell(\cos\theta),$$0. The paper states that $$C(\theta)=\sum_{\ell=0}^\infty C_\ell \frac{2\ell+1}{4\pi}P_\ell(\cos\theta),$$1 is nonzero at about the 95% confidence level (Agazie et al., 2024). When higher multipoles are added, the Bayes factors drop from approximately $$C(\theta)=\sum_{\ell=0}^\infty C_\ell \frac{2\ell+1}{4\pi}P_\ell(\cos\theta),$$2 for $$C(\theta)=\sum_{\ell=0}^\infty C_\ell \frac{2\ell+1}{4\pi}P_\ell(\cos\theta),$$3 to approximately $$C(\theta)=\sum_{\ell=0}^\infty C_\ell \frac{2\ell+1}{4\pi}P_\ell(\cos\theta),$$4 for $$C(\theta)=\sum_{\ell=0}^\infty C_\ell \frac{2\ell+1}{4\pi}P_\ell(\cos\theta),$$5 and approximately $$C(\theta)=\sum_{\ell=0}^\infty C_\ell \frac{2\ell+1}{4\pi}P_\ell(\cos\theta),$$6 for $$C(\theta)=\sum_{\ell=0}^\infty C_\ell \frac{2\ell+1}{4\pi}P_\ell(\cos\theta),$$7, indicating that the data strongly prefer a model that is essentially quadrupolar (Agazie et al., 2024).

Earlier harmonic-space work had already shown why this outcome is structurally natural. For a statistically isotropic Gaussian background, the redshift-map power spectrum is

$$C(\theta)=\sum_{\ell=0}^\infty C_\ell \frac{2\ell+1}{4\pi}P_\ell(\cos\theta),$$8

and PTA sensitivity to the angular power spectrum is overwhelmingly dominated by the quadrupole anisotropy: $$C(\theta)=\sum_{\ell=0}^\infty C_\ell \frac{2\ell+1}{4\pi}P_\ell(\cos\theta),$$9 contributes $$\Gamma^{\rm HD}_{ab}=(1+\delta_{ab})\sum_{\ell=2}^{\infty} c_\ell^{\rm HD} P_\ell(\cos\theta_{ab}),$$0 of $$\Gamma^{\rm HD}_{ab}=(1+\delta_{ab})\sum_{\ell=2}^{\infty} c_\ell^{\rm HD} P_\ell(\cos\theta_{ab}),$$1, equivalently measuring only the quadrupole gives about $$\Gamma^{\rm HD}_{ab}=(1+\delta_{ab})\sum_{\ell=2}^{\infty} c_\ell^{\rm HD} P_\ell(\cos\theta_{ab}),$$2 of the full signal-to-noise (Roebber et al., 2016). In this sense, the dominance of $$\Gamma^{\rm HD}_{ab}=(1+\delta_{ab})\sum_{\ell=2}^{\infty} c_\ell^{\rm HD} P_\ell(\cos\theta_{ab}),$$3 is not merely an empirical feature of one data release but a consequence of the red spectrum of the angular response.

The same formalism also clarifies source-correlation effects beyond the mean Hellings–Downs curve. Rotationally invariant ensembles built from anisotropic Gaussian subensembles can preserve the mean HD curve while changing the cosmic covariance and total covariance through an angular covariance $$\Gamma^{\rm HD}_{ab}=(1+\delta_{ab})\sum_{\ell=2}^{\infty} c_\ell^{\rm HD} P_\ell(\cos\theta_{ab}),$$4 with multipoles $$\Gamma^{\rm HD}_{ab}=(1+\delta_{ab})\sum_{\ell=2}^{\infty} c_\ell^{\rm HD} P_\ell(\cos\theta_{ab}),$$5 (Allen, 2024). A practical implication is that “statistical isotropy” of the ensemble does not imply that any given realization is isotropic; harmonic-space covariance is the appropriate place where that distinction becomes operational (Allen, 2024).

3. Cosmological sky fields, pseudo-spectra, and direct harmonic estimators

In cosmology, angular harmonic space analysis is used both for signal separation and for power-spectrum estimation. In 21 cm intensity mapping, the observed map is written as

$$\Gamma^{\rm HD}_{ab}=(1+\delta_{ab})\sum_{\ell=2}^{\infty} c_\ell^{\rm HD} P_\ell(\cos\theta_{ab}),$$6

and then expanded into spherical harmonics,

$$\Gamma^{\rm HD}_{ab}=(1+\delta_{ab})\sum_{\ell=2}^{\infty} c_\ell^{\rm HD} P_\ell(\cos\theta_{ab}),$$7

At each multipole, this yields an empirical frequency-frequency covariance matrix

$$\Gamma^{\rm HD}_{ab}=(1+\delta_{ab})\sum_{\ell=2}^{\infty} c_\ell^{\rm HD} P_\ell(\cos\theta_{ab}),$$8

which is the basis of a harmonic-space internal linear combination method informed by the principal components of the theoretical 21 cm covariance (Joseph et al., 2024). The weights

$$\Gamma^{\rm HD}_{ab}=(1+\delta_{ab})\sum_{\ell=2}^{\infty} c_\ell^{\rm HD} P_\ell(\cos\theta_{ab}),$$9

minimize variance while preserving the signal subspace (Joseph et al., 2024). The full-sky angular power spectrum is then reconstructed from masked maps through the mode-coupling matrix $$c_\ell^{\rm HD}=\frac{3}{2}(2\ell+1)\frac{(\ell-2)!}{(\ell+2)!},\qquad \ell\ge 2,$$0 (Joseph et al., 2024).

For projected large-scale structure traced by galaxies, harmonic-space analysis need not begin with a pixelized map. The harmonic coefficients can be computed directly from a weighted point set,

$$c_\ell^{\rm HD}=\frac{3}{2}(2\ell+1)\frac{(\ell-2)!}{(\ell+2)!},\qquad \ell\ge 2,$$1

which avoids aliasing of small-scale power and pixel window functions (Lizancos et al., 2023). The pseudo-spectrum obeys the standard coupling relation

$$c_\ell^{\rm HD}=\frac{3}{2}(2\ell+1)\frac{(\ell-2)!}{(\ell+2)!},\qquad \ell\ge 2,$$2

with the window spectrum $$c_\ell^{\rm HD}=\frac{3}{2}(2\ell+1)\frac{(\ell-2)!}{(\ell+2)!},\qquad \ell\ge 2,$$3 entering through Wigner $$c_\ell^{\rm HD}=\frac{3}{2}(2\ell+1)\frac{(\ell-2)!}{(\ell+2)!},\qquad \ell\ge 2,$$4 symbols (Lizancos et al., 2023). The direct point-catalog approach is mathematically equivalent to the map-based pseudo-$$c_\ell^{\rm HD}=\frac{3}{2}(2\ell+1)\frac{(\ell-2)!}{(\ell+2)!},\qquad \ell\ge 2,$$5 method when the number of pixels is sufficiently large and the mask is well sampled, but it eliminates pixelization artifacts at finite resolution (Lizancos et al., 2023).

A related development concerns redshift-space distortions in nearly full-sky spectroscopic surveys. A naive projection of higher-order RSD terms into harmonic space leads to divergent Bessel-integral expressions, but the nonlinear RSD effect, including the fingers-of-God, can be entirely attributed to a modification of the radial window function (Gebhardt et al., 2020). The practical criterion

$$c_\ell^{\rm HD}=\frac{3}{2}(2\ell+1)\frac{(\ell-2)!}{(\ell+2)!},\qquad \ell\ge 2,$$6

marks the regime where FoG suppress the angular power spectrum on all transverse scales (Gebhardt et al., 2020). The paper also provides a flat-sky approximation that reproduces the full calculation to sub-percent accuracy (Gebhardt et al., 2020). This is a characteristic example of harmonic-space analysis turning a formally problematic expansion into a well-defined window-function problem.

4. Orbital angular momentum, spatiotemporal vortices, and harmonic generation

In nonlinear optics, angular harmonic space analysis appears as the decomposition of fields into OAM or vorticity channels. For ideal Laguerre–Gauss beams, the familiar rule

$$c_\ell^{\rm HD}=\frac{3}{2}(2\ell+1)\frac{(\ell-2)!}{(\ell+2)!},\qquad \ell\ge 2,$$7

coincides with orbital angular momentum conservation because the topological charge and OAM per photon satisfy $$c_\ell^{\rm HD}=\frac{3}{2}(2\ell+1)\frac{(\ell-2)!}{(\ell+2)!},\qquad \ell\ge 2,$$8. For generic structured drivers, however, topological charge and OAM per photon can decouple, and the universal scaling law is instead

$$c_\ell^{\rm HD}=\frac{3}{2}(2\ell+1)\frac{(\ell-2)!}{(\ell+2)!},\qquad \ell\ge 2,$$9

or, in the undepleted-driver HHG limit,

$c_0^{\rm HD}=c_1^{\rm HD}=0$0

The central claim is that what scales with harmonic order is the OAM converted per photon converted, not the input driver’s OAM per photon in an absolute sense (Porras et al., 9 Mar 2026).

This distinction is especially important for spatiotemporal optical vortices. In second-harmonic generation of ST-OAM pulses, the ideal nonlinear map is

$c_0^{\rm HD}=c_1^{\rm HD}=0$1

and experimental angular decomposition on the $c_0^{\rm HD}=c_1^{\rm HD}=0$2 plane shows that the fundamental is dominated by $c_0^{\rm HD}=c_1^{\rm HD}=0$3 while the second harmonic is dominated by $c_0^{\rm HD}=c_1^{\rm HD}=0$4 (Gui et al., 2021). The reported modal powers are approximately $c_0^{\rm HD}=c_1^{\rm HD}=0$5 in $c_0^{\rm HD}=c_1^{\rm HD}=0$6 for the fundamental, $c_0^{\rm HD}=c_1^{\rm HD}=0$7 in $c_0^{\rm HD}=c_1^{\rm HD}=0$8 for thin-BBO SHG, and $c_0^{\rm HD}=c_1^{\rm HD}=0$9 in $\ell$0 for thick-BBO SHG, with a significant DC component in the thick-crystal case (Gui et al., 2021). The same literature emphasizes that charge conservation or scaling is robust, whereas topological form and mode purity are not guaranteed because phase mismatch, group velocity mismatch, and dispersion can split or distort singularities (Gui et al., 2021, Hancock et al., 2020).

Structured-light HHG extends the same angular-harmonic logic to transverse OAM. STOV-driven HHG produces spatially resolved harmonic spectra with tilted harmonic ridges and fine interference patterns, and the $\ell$1-th harmonic inherits transverse topological charge $\ell$2, with numerical values such as $\ell$3, $\ell$4, $\ell$5, and $\ell$6 for $\ell$7 (Fang et al., 2023). In a two-color counter-spin and counter-vorticity scheme, the $\ell$8-th harmonic is forbidden and each surviving harmonic is associated with a unique channel, so the transverse OAM and SAM are simultaneously fixed by selection rules (Fang et al., 2023).

Other nonlinear optical realizations highlight how angular harmonic space analysis couples azimuthal and radial mode content. In type II SHG with orthogonally polarized Laguerre–Gaussian inputs, the overlap integral yields the OAM selection rule

$\ell$9

while opposite-sign input helicities excite radial modes up to

$\ell^{-3}$0

Thus the nonlinear crystal acts as a selector of both angular and radial harmonic content, not only of total topological charge (Pereira et al., 2017). Related experiments on spin-constrained HHG show that local polarization structure can restrict allowed multiphoton channels so that the emitted harmonic OAM is locked to the allowed microscopic spin channels (Kong et al., 2018), while relativistic interaction of a circularly polarized beam with a plane foil shows spin-to-orbital conversion driven by total angular momentum conservation (Li et al., 2018).

5. Tensorial and group-theoretic formulations

A tensorial version of angular harmonic space analysis replaces scalar harmonics by symmetric traceless tensors. The rank-$\ell^{-3}$1 Cartesian harmonic tensor $\ell^{-3}$2 is built from $\ell^{-3}$3 copies of a unit vector and projected onto the completely symmetric traceless subspace (Parke, 2023). Low-rank examples make the construction explicit: $\ell^{-3}$4 and higher ranks are obtained by systematic trace subtraction (Parke, 2023). The framework is useful for expressing couplings of multiple spherical harmonics entirely through dot products of unit vectors, and an assembly-language implementation generated arbitrary-rank tensors efficiently; for the $\ell^{-3}$5 verification case, Mathematica took about $\ell^{-3}$6 seconds on a 386-i7 PC while the compiled assembly program took less than $\ell^{-3}$7 ms (Parke, 2023).

A group-theoretic generalization appears in the construction of harmonic wave functions for integer and half-integer angular momentum on $\ell^{-3}$8. Using the Hurwitz–Hopf map and Schwinger’s two-dimensional harmonic oscillators, the framework lifts ordinary $\ell^{-3}$9 dependence to $$F(\Omega;\Omega_p)=\sum_{l=2}^{\infty}\sum_{m=-l}^{l} A_l\,{}_2Y_{lm}(\Omega)\,Y_{lm}(\Omega_p), \qquad A_l=\frac{1}{\sqrt{(l+2)(l+1)l(l-1)}},$$0, so that spinorial harmonics carry the additional $$F(\Omega;\Omega_p)=\sum_{l=2}^{\infty}\sum_{m=-l}^{l} A_l\,{}_2Y_{lm}(\Omega)\,Y_{lm}(\Omega_p), \qquad A_l=\frac{1}{\sqrt{(l+2)(l+1)l(l-1)}},$$1 fiber angle $$F(\Omega;\Omega_p)=\sum_{l=2}^{\infty}\sum_{m=-l}^{l} A_l\,{}_2Y_{lm}(\Omega)\,Y_{lm}(\Omega_p), \qquad A_l=\frac{1}{\sqrt{(l+2)(l+1)l(l-1)}},$$2 (Hojman et al., 2022). The operators

$$F(\Omega;\Omega_p)=\sum_{l=2}^{\infty}\sum_{m=-l}^{l} A_l\,{}_2Y_{lm}(\Omega)\,Y_{lm}(\Omega_p), \qquad A_l=\frac{1}{\sqrt{(l+2)(l+1)l(l-1)}},$$3

satisfy the usual angular-momentum algebra, while additional oscillator-generated operators change total angular momentum by half a unit (Hojman et al., 2022). This suggests that angular harmonic space analysis can be formulated not only on $$F(\Omega;\Omega_p)=\sum_{l=2}^{\infty}\sum_{m=-l}^{l} A_l\,{}_2Y_{lm}(\Omega)\,Y_{lm}(\Omega_p), \qquad A_l=\frac{1}{\sqrt{(l+2)(l+1)l(l-1)}},$$4 but on its double cover when half-integer representations are essential.

6. Misconceptions, gauge issues, and unresolved questions

Several papers emphasize that angular harmonic space analysis clarifies, rather than removes, physically important ambiguities. In PTA response theory, the phase of the response depends on the choice of polarization basis and is therefore gauge dependent, although the modulus and observable covariances are gauge invariant (Allen, 2024). The diagonalized harmonic form isolates this dependence cleanly into spin-weighted phases.

In the NANOGrav 15-year analysis, a major caveat is the unexplained monopolar common-spectrum feature near $$F(\Omega;\Omega_p)=\sum_{l=2}^{\infty}\sum_{m=-l}^{l} A_l\,{}_2Y_{lm}(\Omega)\,Y_{lm}(\Omega_p), \qquad A_l=\frac{1}{\sqrt{(l+2)(l+1)l(l-1)}},$$5 nHz. When a monopole free-spectrum component is added alongside the quadrupole model, the Savage–Dickey Bayes factor for $$F(\Omega;\Omega_p)=\sum_{l=2}^{\infty}\sum_{m=-l}^{l} A_l\,{}_2Y_{lm}(\Omega)\,Y_{lm}(\Omega_p), \qquad A_l=\frac{1}{\sqrt{(l+2)(l+1)l(l-1)}},$$6 drops from about $$F(\Omega;\Omega_p)=\sum_{l=2}^{\infty}\sum_{m=-l}^{l} A_l\,{}_2Y_{lm}(\Omega)\,Y_{lm}(\Omega_p), \qquad A_l=\frac{1}{\sqrt{(l+2)(l+1)l(l-1)}},$$7 to about $$F(\Omega;\Omega_p)=\sum_{l=2}^{\infty}\sum_{m=-l}^{l} A_l\,{}_2Y_{lm}(\Omega)\,Y_{lm}(\Omega_p), \qquad A_l=\frac{1}{\sqrt{(l+2)(l+1)l(l-1)}},$$8, showing that allowing $$F(\Omega;\Omega_p)=\sum_{l=2}^{\infty}\sum_{m=-l}^{l} A_l\,{}_2Y_{lm}(\Omega)\,Y_{lm}(\Omega_p), \qquad A_l=\frac{1}{\sqrt{(l+2)(l+1)l(l-1)}},$$9 substantially reduces the apparent strength of quadrupolar evidence (Agazie et al., 2024). The paper states that clock errors do not currently explain this feature, and mentions possible explanations including a single supermassive black hole binary, ultralight dark matter, nonstandard GW polarization modes, or some other astrophysical or cosmological effect, but concludes that none of these explanations has yet been established (Agazie et al., 2024).

In structured-light harmonic generation, the literature repeatedly warns that topological charge is not a universal proxy for OAM. A beam can have vanishing TC but nonzero OAM, or nonzero TC and zero OAM, and spatiotemporal optical vortices may change TC during free propagation while their OAM remains constant (Porras et al., 9 Mar 2026). Likewise, in STOV SHG the ideal $l,m$0 rule may coexist with splitting of a nominal $l,m$1 structure into two separated $l,m$2 singularities because of group velocity mismatch and group delay dispersion (Hancock et al., 2020). The conserved quantity is then the total winding or total OAM, not necessarily a single-core topology.

A broader implication is that angular harmonic space analysis is most reliable when the basis is chosen to match the conserved or observationally relevant quantity: Legendre multipoles for isotropic pair correlations, spin-weighted harmonics for tensor fields on the sphere, direct spherical harmonics for pseudo-spectra of discrete tracers, and longitudinal, transverse, or intrinsic transverse OAM channels for nonlinear optics (Joseph et al., 2024, Lizancos et al., 2023, Porras et al., 9 Mar 2026). Where several angular notions coexist, the analysis is most informative when it distinguishes them explicitly rather than treating them as interchangeable.