The Hidden Geometry of Astrophysical Spectra: Path-Signatures of Line Profiles The Hidden Geometry of Astrophysical Spectra: Path-Signatures of Line Profiles – A.2
Rafael S. de Souza İD, Severin Bunk İD · 2026
Read The Hidden Geometry of Astrophysical Spectra: Path-Signatures of Line Profiles The Hidden Geometry of Astrophysical Spectra: Path-Signatures of Line Profiles – A.2 by Rafael S. de Souza İD, Severin Bunk İD (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-27432.txt.
Introduction
de Souza and Bunk read the shape of a spectral line the way one reads a path, not a list of summary numbers. Standard summaries cannot distinguish broad wings, shoulders, double peaks, or layered emission and absorption. Borrowing from rough path theory, each profile becomes a trajectory across a velocity-flux plane, swept from blue to red. A compact set of low-order measures then captures signed area, blue-red imbalance, shape complexity, and the ordering of emission against absorption in this astrophysics paper from 2026.
Synthetic profiles show the measures separating shapes that share the same width and moment summaries. Applied to integral-field spectroscopy of galaxies, the measures are computed pixel by pixel and clustered. The resulting classes form coherent regions of similar line shape. Stacked spectra within each region recover large-scale velocity patterns matching reference velocity fields, although no velocity information was given to the clustering.
Catalog record
| CREATOR | Rafael S. de Souza İD, Severin Bunk İD |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 85 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.27432
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
The Hidden Geometry of Astrophysical Spectra: Path-Signatures of Line Profiles The Hidden Geometry of Astrophysical Spectra: Path-Signatures of Line Profiles – A.2 Rafael S. de Souza İD, Severin Bunk İD Abstract The morphology of a spectral-line profile contains information beyond scalar summaries of line strength, centroid, width, global asymmetry, or diagnostic line ratios. Broad wings, shoulders, double peaks, secondary components, and composite emission–absorption structures encode how flux is ordered across wavelength but can remain indistinguishable under conventional summaries. We introduce an interpretable geometric representation of line profiles inspired by rough path theory. Each wavelength-sampled profile is mapped to a common systemic rest-frame velocity grid and treated as a trajectory in velocity–flux space, traversed from blue to red. From this path, we define a compact set of low-order descriptors measuring signed velocity–flux area, blue–red imbalance localization, higher-order shape complexity, and emission–absorption ordering. Using synthetic profiles, we show that these descriptors separate morphologies with similar full width at half maximum (FWHM), non-parametric velocity width ( W 80 W_{80} ), and low-order moment summaries. We then apply the method to MaNGA integral-field spectroscopy by computing H α \alpha descriptors in individual spaxels and clustering them in a low-dimensional feature space. The resulting classes form spatially coherent regions of similar ordered line morphology. Although no external velocity field is supplied to the clustering, stacked spectra within these regions recover coherent large-scale centroid-velocity patterns broadly consistent with the MaNGA reference velocity fields. We release a minimalist MIT-licensed package spectropath , available at the project website . 1 Introduction Scientific progress often depends on finding representations that make the relevant structure of a problem explicit ( 17 ) . Fourier expansions made signals amenable in frequency space ( 16 ) , Bayesian probability provided a calculus for uncertainty ( 3 ; 12 ) , Hamiltonian and Lagrangian mechanics reformulated dynamical evolution ( 24 ; 19 ) , differential geometry gave a natural language for general relativity ( 15 ) , and information theory reframed communication in terms of entropy and coding ( 39 ) . The common lesson is that a representation is not merely a change of notation, but rather defines which aspects of a system become easy to measure, compare, compress, and interpret. In astronomy, task-aware compression appears in several related but distinct forms. Spectroscopic and photometric surveys have long been represented through Karhunen–Loève or principal-component projections, which reduce high-dimensional data by retaining directions of dominant variance ( 41 ; 10 ; 11 ; 47 ; 48 ) . A more explicitly inferential form of compression is used for galaxy spectra, where optimized linear summaries can be constructed to preserve the Fisher information relevant to a chosen set of physical parameters, rather than merely preserving variance ( 20 ; 35 ; 2 ) . In the ideal case of parameter-independent noise covariance, a spectrum with thousands of wavelength pixels can be compressed into one number per parameter without loss of Fisher information, making the compressed representation directly tied to model sensitivities such as age, normalization, metallicity, or star-formation history. Other astronomical data products are compressed through sparse and multiscale decompositions, which preserve localized image structure across position and scale ( 40 ) , or through shapelet bases, which encode galaxy morphology in a compact set of analytic coefficients ( 34 ) . Related representations include stellar-population synthesis and spectral decomposition ( 9 ; 38 ) , equivalent widths and Lick-style absorption indices ( 46 ; 42 ) , as well as machine-learning methods for denoising and feature extraction ( 5 ; 45 ) . In this sense, compression should be understood not only as a reduction in data volume, but as a choice of representation. Its value lies in preserving those aspects of the data that are informative for a given scientific question, while suppressing degrees of freedom that are incidental to that question. Here we follow the same guiding principle in the context of spectral-line morphology. A spectral-line profile contains more than an integrated flux, centroid, width, or global asymmetry: it is an ordered function of wavelength, or equivalently of line-of-sight velocity, in which emission and absorption occur at specific positions relative to the systemic frame. This ordering is physically meaningful because different mechanisms imprint structure in different parts of the profile. Blue-shifted interstellar absorption in NaI D-doublet lines is a classical signature of starburst-driven outflowing material along the line of sight ( 21 ; 44 ) , while broad and blueshifted [O iii ] components are commonly used to trace ionized outflows in active galaxies ( 31 ) . Double-peaked narrow emission lines make this loss of ordered structure especially clear. A two-component profile is not simply a broader line: the relative placement and strength of the blue and red components indicate that the flux is distributed between distinct velocity channels, although similar unresolved profiles may arise from disk rotation, narrow-line-region outflows, dual active nuclei ( 32 ) , or biconical flows ( 33 ) , and double-peaked signatures can also be produced by the central rotation curve and bars in disc galaxies ( 29 ) . Mixed emission–absorption profiles make the role of ordering especially explicit: P-Cygni profiles, with blueshifted absorption and redshifted emission, are associated with outflowing material ( 37 ) , whereas inverse P-Cygni profiles or redshifted absorption are used as evidence for inflowing gas ( 22 ) . Thus, two profiles with similar flux, width, or scalar asymmetry can still encode different physical configurations if the sequence of blue-side absorption, blue-side excess, core emission, red-side excess, and absorption differs across velocity. Quantities such as line strength, centroid velocity, velocity dispersion, percentile widths, asymmetry measures, and Gauss–Hermite coefficients provide compact and physically useful summaries of spectra and line-of-sight velocity distributions ( 43 ; 6 ) . Our aim is complementary: to construct descriptors that preserve the ordered morphology of individual line profiles. To achieve this, we will employ path signatures. Path signatures provide a natural mathematical representation for ordered data. Originating in the theory of iterated integrals and rough paths ( 7 ; 27 ) , the signature of a path maps an ordered curve into a hierarchy of tensor coordinates. These coordinates record not only the net displacement of the curve, but also the precise sequence in which its coordinate increments are accumulated. Under suitable regularity conditions, the full signature provides a highly informative description of an ordered curve while remaining invariant under reparameterization ( 4 ) . We formalize this by treating a continuum-subtracted line profile not merely as a discrete vector of flux samples (F_{1},F_{2},\ldots,F_{n}) , but as a directed path, X(v)=(v,F(v)) , traversed continuously from blue to red velocity in the velocity–flux plane. In this framework, wings, shoulders, and complex emission–absorption structures become geometric features of a directed curve. Because path signature coefficients explicitly record the sequence of variations, a blue-side feature followed by a red-side feature yields a fundamentally distinct representation from the reverse configuration. While existing software such as iisignature ( 36 ) , signatory ( 23 ) , esig ( 28 ) , and RoughPy ( 30 ) provide off-the-shelf implementations of path signatures, the accompanying package developed for this work, spectropath , occupies a more specialized locus in this ecosystem. Instead