Exploring the long-term temporal variability in polarization through multi-epoch optical spectro-polarimetry - Part I: A sample of Herbig Ae/Be and classical Be stars

This paper reports a multi-epoch spectro-polarimetric monitoring campaign targeting a sample of Herbig Ae/Be and classical Be stars, conducted using the ProtoPol instrument, a medium-resolution echelle spectro-polarimeter mounted on the Physical Research Laboratory (PRL) 2.5m telescope at Mt. Abu, India. The observations span over 28 months (December 2023–March 2026), with the initial observations obtained during the commissioning and performance-verification phase of ProtoPol. The sample comprises 11 Herbig Ae/Be stars and 10 classical Be stars.

The study found that while the Hα polarization remained relatively constant for the classical Be stars, they showed significant variability for most of the Herbig stars in the sample. The observations constitute one of the rare multi-epoch spectro-polarimetric datasets spanning more than two years.

For the Herbig Ae/Be stars, the multi-epoch observations revealed strong polarization variability across the Hα emission feature, even when the intensity spectrum did not show much change. From the observation sample, 10 out of the 11 Herbig stars showed detectable spectro-polarimetric signatures, at least for one epoch of observation. Most sources showed strong polarization variability over a time span of 12-16 months, except for GU CMa and MWC 442, which did not show any polarization signature across the Hα emission line in both epochs. The observed spectro-polarimetric signatures showed very different polarization morphologies, even for sources with similar intensity spectra. In many of the stars, the strongest observed polarization change is noticed across the absorption component in the Hα emission line, while the polarization across the emission parts of the line is almost at the level of the continuum polarization. In many of the 'windy' systems, which are systems where the Hα emission originated from a circumstellar wind as opposed to 'disky' systems where the line origin is a circumstellar disk, the clearest spectro-polarimetric variability was observed, including sources like MWC 120, MWC 480, and MWC 758, where the polarization change is detected across the central absorption/blue-shifted P-Cygni absorption. In 'disky' systems like MWC 158, a difference in the polarization amplitude is evident between blue and red peaks. The other 'disky' systems in the sample, like HD 58647, MWC 147, and FS CMa, also show strong polarization variability, with the strongest polarization detected across the absorption trough, but the blue and red emission peaks also show polarization above the continuum level, resulting in an apparent broad spectro-polarimetric signature across the Hα emission line.

For the classical Be stars, the multi-epoch spectro-polarimetry, spanning a period of 28 months, provides a good reference sample to demonstrate the accuracy of the obtained results for the Herbig star spectro-polarimetric sample, due to their more predictable Hα polarization behavior. Half of the stars in the sample show depolarization signatures across the Hα emission line. The polarization signatures in these stars are less complicated than those in the Herbig sample, either a depolarization signature or no polarization change. Multi-epoch observations further demonstrate that the Hα polarization signatures in these stars are a lot more constant as compared to their Herbig counterparts, which showed a far greater temporal variability, both in line profiles and polarization. In a few of the stars, even though there may be slight changes in the amplitude of the polarization signatures, the overall line effect remained constant.

The observations reveal a clear variability of the polarization amplitude and polarization profile across the Hα emission for several stars in the sample, especially in the Herbig sample. This includes changes in the intrinsic polarization amplitude or depolarization depth, changes in position angles, and emergence/disappearance of complex line effects. These variations clearly indicate that the scattering geometries and physical conditions within the circumstellar environment are not static, but rather evolve on observable timescales.

For classical Be stars, the variability is broadly consistent, with the targets showing either no polarization signature or a depolarization signature, which remained constant over the observation epochs. On the other hand, the sample of Herbig Ae/Be stars shows diverse spectro-polarimetric effects such as depolarization, intrinsic polarization, McLean effects, or no changes in polarization across the Hα. The polarization signatures are also a lot more variable, even on yearly timescales, not just in amplitude but also in polarization profiles, thus demonstrating a lot more complex scattering environments, likely influenced by a combination of gaseous disks, dust, and ongoing accretion or outflow processes.

The paper is Part-I of a two-part series of sample studies; corresponding results for symbiotic and red giant stars are presented in Part-II.

Improvements to AI Systems:

1. Temporal Variability Prediction for Circumstellar Environments

• Train a model on multi-epoch spectro-polarimetric data (Hα line profiles + polarization) to predict the evolution of scattering geometries in Herbig Ae/Be vs. classical Be stars.

• The AI can classify whether a system is windy or disky based on polarization morphology changes over 12–16 months, even when intensity spectra remain static.

2. Automated Detection of Polarization Morphology Types

• Develop a classifier that distinguishes depolarization, intrinsic polarization, McLean effects, and no-change signatures across Hα, using the observed diversity in Herbig stars.

• The AI can flag rare transient events (e.g., emergence/disappearance of complex line effects) that indicate non-static circumstellar conditions.

3. Cross-Instrument Calibration and Noise Modeling

• Use the classical Be star sample (constant polarization) as a control to calibrate systematic errors in medium-resolution echelle spectro-polarimeters.

• The AI can learn instrument-specific noise patterns and correct for them in future observations, improving signal-to-noise for faint polarization features.

4. Physical Parameter Inference from Polarization Variability

• Build an inverse model that maps observed polarization changes (amplitude, position angle, profile shape) to physical parameters like disk density, wind outflow rate, or dust scattering asymmetry.

• The AI can estimate timescales of geometric evolution (e.g., disk warping, wind clumping) from yearly cadence data.

5. Multi-Epoch Data Fusion for Rare Object Identification

• Train a transformer-based model on the 28-month dataset to identify stars with anomalous polarization behavior (e.g., GU CMa and MWC 442 showing no signatures) and predict which new targets are likely to exhibit strong variability.

• The AI can prioritize follow-up observations for time-critical events in Herbig systems.

6. Cross-Wavelength Generalization

• Extend the model to predict polarization behavior in other emission lines (e.g., Hβ, Paβ) or across broader spectral ranges, using the Hα patterns as a training prior.

• The AI can then be applied to upcoming surveys (e.g., LSST, WEAVE) to pre-screen candidates for detailed spectro-polarimetric study.

7. Simulation-to-Observation Transfer Learning

• Use the observed variability to refine radiative transfer simulations of circumstellar disks and winds.

• The AI can generate synthetic multi-epoch spectro-polarimetric data for training, then fine-tune on real observations to improve predictive accuracy for unseen stars.

What the Improved AI System Can Do:

• Given a single-epoch Hα spectrum and polarization profile, predict the likely range of polarization variability over the next 1–2 years for a Herbig Ae/Be star.

• Automatically classify a star as windy vs. disky with >90% accuracy, even when intensity profiles are ambiguous.

• Detect subtle depolarization changes in classical Be stars that indicate early disk dissipation, useful for monitoring stellar evolution.

• Provide real-time alerts for transient polarization events (e.g., new P-Cygni absorption) in ongoing monitoring campaigns.

• Serve as a virtual assistant for astronomers, suggesting optimal observation cadence and epochs to capture maximum variability for a given target.

Polarization signatures across emission line features, together with their temporal evolution, offer a powerful probe of the circumstellar environments of astrophysical sources, on spatial scales otherwise inaccessible to direct imaging techniques. However, the photon-hungry nature of spectro-polarimetry has significantly limited the availability of such datasets in the literature. This work reports a multi-epoch spectro-polarimetric monitoring campaign targeting a sample of Herbig Ae/Be and classical Be stars. The initial observations were obtained during the commissioning and performance-verification phase of ProtoPol, a recently developed medium-resolution echelle spectro-polarimeter mounted on the Physical Research Laboratory (PRL) 2.5m telescope, Mt Abu, India. Given the limited number of comparable datasets available for this class of objects, follow-up observations of the same targets were carried out repeatedly over more than 28 months (December 2023-March 2026), enabling an investigation of the temporal behavior of their polarimetric properties. Our sample comprises 11 Herbig Ae/Be stars and 10 classical Be stars. Our observations found that, while the H alpha polarization remained relatively constant for the classical Be stars, they showed significant variability for most of the Herbig stars in the sample. The observations presented here constitute one of the rare multi-epoch spectro-polarimetric datasets spanning more than two years and should be of considerable interest to the broader astronomical community. This paper is Part-I of a two-part series of sample studies; corresponding results for symbiotic and red giant stars are presented in Part-II.