TiDES (Time Domain Extragalactic Survey) is a survey on the 4-metre Multi-Object Spectrograph Telescope (4MOST), focused on the spectroscopic follow-up of Rubin Observatory LSST extragalactic optical transients. We have 250,000 fibre-hours of spectroscopy time available within the TiDES survey (2% of the total 4MOST fibres). With this, we expect to obtain:

1. Spectroscopic observations of > 30,000 live transients to r_AB = 22.5.
2. Follow up observations of > 200,000 transient host galaxies to obtain redshift measurements for photometric classification and cosmological applications.

We expect there to be between 5–10 live (i < 22.5 mag) extragalactic transients per 4MOST field, and we will put a fibre on all of these, effectively giving a magnitude-limited sample.

This project focuses on providing non-directable software development effort to ensure the maximal science return from the TiDES dataset for the Rubin Observatory science consortia. This includes using the resulting dataset for supernova cosmology in the DESC (primarily via type Ia supernovae in the Time Domain working group), and understanding supernova physics in the TVS science consortium. The project will thus support developing TiDES into a contributed dataset to DESC and TVS.

We will work closely with the DESC Time Domain Working Group to design a selection function for the LSST transient stream to optimise the type Ia SN selection (including host galaxies) for cosmology. We will implement the DESC agreed strategy within the Lasair broker to select supernovae with high-quality LSST light curves (or other measured characteristics).

The integrated development of Quasar Neural Process in Python (QNPy) and Quasar harmonic eXplorer (QhX) delivers a framework for the nonlinear, probabilistic analysis of quasar light curves, addressing the challenges of irregular sampling, noise, and multimodal variability in large time-domain surveys like the Vera C. Rubin LSST. QNPy combines Self-Organizing Maps (SOMs) for clustering with Attentive Latent Neural Processes to reconstruct light curves in a space-context-aware latent representation. In this latent space, Multi-Dimensional Modeling (MDM) infers supermassive black hole (SMBH) mass, transfer functions, and characteristic variability time scales.

QhX introduces a novel period–period phase space by transforming light curves using wavelets and correlating the resulting time–frequency representations. QhX introduces a new application of the Intersection over Union (IoU) metric to statistically validate periodic detections across photometric bands, and both numerical outputs and dynamic visualizations of detected periods, supporting robust classification of multiperiodic signals. A statistical robovetter determines the significance and confidence bounds of detected periods, producing a catalog of reliable periodic quasar candidates. This method enables the detection of complex periodic patterns via a nonlinear correlation density map. The QNPy–QhX framework is modular and interoperable, designed to operate in tandem or be combined with other pipelines and methodologies. This flexibility enables its integration into ensemble workflows for quasar classification, variability characterization, and periodicity detection.

The expected outcomes include:

1. Catalogues of binary quasar candidates suitable for follow-up by future observatories.
2. Modeled quasar light curves reconstructed with quantified uncertainties.
3. Inference of transfer functions, SMBH mass, and characteristic time scales of variability from real and simulated quasar light curves.

We proposed the development of a machine-learning based classification pipeline to incrementally improve classification of periodic variable sources with LSST multicolor light curves as more and more data are coming in during the 10-yr main survey. Since a monolithic classifier would not be able to meet the output purity requirements we propose an iterative modular system with different modules tuned for classification of specific transients and variable stars. We would take advantage of the fact that some of the modules may already be in development in other LSST Science Collaborations and incorporate these when possible in the proposed pipeline. The pipeline will be primarily based on LSST multicolor lightcurves, but would also take advantage of other LSST data products where possible (e.g. real/bogus brokers) as well as data from other surveys.

The product to be developed is a Target and Observing Program Management system web-service that will allow TVS members to manage and track active observing programs. The system will be designed to handle observing request management, alert tracking and data administration.

We propose to work within the Crowded Fields Working Group, and in agreement with the SMWLV and TVS Science Collaborations, in the framework of Directable Software, to implement a series of tools focused on obtaining accurate photometry in crowded fields.

As a proposed example, we mention our experience in the P.B. Stetson’s ALLFRAME code, which simultaneously fits a master star list on a set of images, cross-correlating fluxes (in different passbands) and positions. This approach allows for enhanced accuracy of the photometry, especially at the faint end of the brightness range. We are available to test this approach and other approaches developed to deal with the LSST data, according to the directable nature of our proposed contribution. In general, we are keen to contribute with specific FTEs and our experience on this matter.

INAF plans to provide directable software effort, using INAF-secured funding, at the level of 0.8 FTE per year for 4 years, starting in FY25. This effort will initially come from a senior postdoc (to be hired) who has software engineering skills and a stellar photometry background, supported by our experienced scientists at a lower (but still dedicated) effort level.

Deliverable products include: (i) software development for PSF-based reduction of crowded fields. The software will be made available to the Community, starting from its development phase, and fully documented and maintained; (ii) photometric catalogs, produced by the aforementioned software. They will be immediately delivered to the Community, before further analysis takes place, in order to avoid any perceived advantage in the subsequent scientific analysis. Software development will take place in collaboration and agreement with the interested Science Collaborations.

On-demand PSF photometry reduction of selected areas is under evaluation.

We propose the development of an infrastructure based on extended model sets along with software tools to interpolate among grids of theoretical templates (e.g., light and radial velocity curves, periods, mean magnitudes, colors, etc.) for pulsating variables. The fundamental aims are:

• To train the developed tools and infrastructure on the extended and detailed model grids for several classes of pulsating stars built by our team.
• To extend the same tools and infrastructure to theoretical LSST pulsating star templates developed by other teams and under the supervision of the TVS Scientific Collaboration.

• Web page access to import and export different empirical and/or theoretical light curve databases for different classes of intrinsic variable stars.
• Interactive software tools to interpolate within the databases and derive the properties of light curves and the associated stellar parameters.
• Interactive software tools to simulate LSST variable stars starting from empirical or theoretical light curve templates.

Catalogues of variable sources of different types will be used to:

1. Intercalibrate the LSST and Gaia datasets using both variable and constant stars within the reach of Gaia and not yet saturated in the LSST.
2. Build training sets for the classification of variable sources observed by LSST.
3. Optimally translate into the LSST passbands the different diagnostic tools developed as part of Coordination Unit 7 (variability) in the Gaia Data Processing and Analysis Consortium (DPAC).
4. Develop and fine-tune various Machine Learning and Deep Learning models for the processing and characterization of RR Lyrae stars and Cepheids.