The Galah survey and symbiotic stars. II. Discovery, validation, and intensive follow-up of 44 new systems accreting at the lowest detectable levels

Introduction

Symbiotic stars (SySts) are interacting binaries composed of a red giant (RG) and a degenerate companion (usually a white dwarf (WD), or very rarely a neutron star), typically orbiting with periods of a few years and separations of a few astronomical units. The companion accretes from the RG via wind interception or Roche-lobe overflow; so far, there is little evidence that magnetic fields play a significant role in this process, except perhaps in a few isolated cases (Sokolowski & Bildsten 1999; Magdolen et al. 2023; Pichardo Marcano et al. 2026).

The classification criteria for SySts have widened over time. The original definition formulated by Allen [6] tended to isolate SySts undergoing nuclear burning at the surface of a white dwarf (e.g., Sokolowski et al. 2001, 2016; Mikołajewska 2003; Munari 2019; Mercè 2025). Later, Belczyński et al. (2000) relaxed these constraints to include systems that are not burning but are accreting onto a WD at a high rate. The discovery

@page 1/26 starts_mid=0 ends_mid=1

Introduction

Symbiotic stars (SySts) are interacting binaries composed of a red giant (RG) and a degenerate companion (usually a white dwarf (WD), or very rarely a neutron star), typically orbiting with periods of a few years and separations of a few astronomical units. The companion accretes from the RG via wind interception or Roche-lobe overflow; so far, there is little evidence that magnetic fields play a significant role in this process, except perhaps in a few isolated cases (Sokolowski & Bildsten 1999; Magdolen et al. 2023; Pichardo Marcano et al. 2026).

The classification criteria for SySts have widened over time. The original definition formulated by Allen [6] tended to isolate SySts undergoing nuclear burning at the surface of a white dwarf (e.g., Sokolowski et al. 2001, 2016; Mikołajewska 2003; Munari 2019; Mercè 2025). Later, Belczyński et al. (2000) relaxed these constraints to include systems that are not burning but are accreting onto a WD at a high rate. The discovery that some SySts contain a neutron star (NS) instead of a WD (e.g., Masetti et al. [87])—along with the recognition of SU Lyn as the prototype of a large, previously unsuspected population of SySts accreting on WDs at (very) low rates (Mukai et al. [101])—has shaped the modern definition; i.e., any binary where a WD or an NS accretes enough material from an RG companion such that the interaction is detectable at some wavelength is considered a SySt.

The traditional view (e.g., Kenyon [73]) implicitly assumed a steady-state evolution for the different types of SySts; thus, a burning system was thought to keep burning for its entire symbiotic lifetime. This posed the problem of refueling the burning shell, which was circumvented by assuming that—regardless of radiation pressure or stellar winds—accretion onto the WD proceeds unabated, matching precisely the rate at which fuel is consumed. In a more modern approach (as discussed by Munari [102]), SySts are seen to alternate between burning phases and accreting-only states; longer intervals are spent accreting material onto the surface of the WD, which is subsequently ignited once the conditions for nuclear burning are reached. When the shell fuel is exhausted, burning ceases and accretion resumes to prepare for a new cycle.

In addition to being fascinating astrophysical laboratories in their own right, SySts have far-reaching implications for a broad range of topics. These include the identification of progenitors for Type Ia supernovae (e.g., Whelan & Iben [140]; Munari & Renzini [106]; Chen et al. [24]; Starrfield et al. [133]; Ruiter & Seitenzahl [123]), the agents responsible for the ultraviolet (UV) excess in elliptical galaxies (e.g., Code [25]; Greggio & Renzini [52]; O’Connell [115]; Chavez & Bertone [22]), the complex mechanisms shaping planetary nebulae (e.g., Corradi et al. [28]; Jones & Boffin [68]; De Marco et al. [34]), and the production–destruction cycle for dust grains under hostile conditions (e.g., Whitelock [141]; Munari et al. [105]; Shore & Gehrz [127]; Jones & Nuth [67]).

The relevance of the symbiotic channel scales with the total number of SySts in the Galaxy, $N(\mathrm{SySts})$, which has been variously estimated in the past (Allen [6]; Kenyon [73]; Munari & Renzini [106]; Kenyon et al. [74]; Yungelson et al. [146]; Magrini et al. 2003; Lü et al. 2006; Laversveiler et al. [79], among others), usually by counting known objects within a given volume and scaling the results to the whole Galaxy. These estimates range over 1,500 $\lesssim N(\mathrm{SySts}) \lesssim$ 500,000, suggesting that we still lack a definitive consensus.

These estimates of $N(\mathrm{SySts})$ rested primarily on the few hundred objects listed in the available catalogs of symbiotic stars (Allen [6]; Belczyński et al. [11]; Merc et al. [91]; Akras et al. [3]), which were dominated by burning systems. Due to the immense luminosity radiated by the burning shell—even exceeding that of the RG—these systems are relatively easy to identify throughout the Galaxy (Allen [6]) and even within the Local Group (Magrini et al. 2003). Far more challenging to detect are the accreting-only SySts, in which accretion processes radiate orders of magnitude less than a burning shell or the RG. As outlined in Appendix § 10, basic evolutionary considerations suggest that SySts spend the (vast) majority of their lifetime in an accreting-only state. Consequently, this subclass likely represents the dominant population for the true $N(\mathrm{SySts})$.

In addition, accretion onto the WD during these non-burning intervals may vary by orders of magnitude between high-accretion (HighAcc) and low-accretion (LowAcc) states, as exemplified by T CrB, AT 2023txn, and SU Lyn (Lopes de Oliveira et al. [80]; Luna et al. [83]; Munari et al. [108, 110]), which spend most of their time in LowAcc states characterized by a weak UV excess, feeble emission lines (usually detectable only in high-resolution spectra), and accretion luminosities of $L_{\mathrm{acc}}\sim10^{-1} L_{\odot}$. At least in T CrB, these low states seem to be corresponding to the gradual buildup of mass in the accretion disk, with little material actually transferred to the central WD. Only when the disk grows beyond stable conditions does it flush most of its content onto the WD in a bright, shorter lived HighAcc episode ($L_{\mathrm{acc}}\sim10^{2} L_{\odot}$) marked by strong UV excess and intense emission lines. If this behavior is common, the majority of accreting-only SySts in the Galaxy are expected to be in a LowAcc state.

In this series of papers, we aim to discover SySts precisely in the LowAcc state. Although they are the most difficult to identify, they represent the most abundant population in the Galaxy and are therefore the most reliable tracers to constrain $N(\mathrm{SySts})$. To this end, we targeted non-pulsating RGs identified in the high-resolution spectra collected by the Galah survey of the southern sky (De Silva et al. [35]; Kos et al. [75]; Buder et al. [18, 17]) that exhibit even minimal H$\alpha$ emission. We subjected these candidates to extensive multiwavelength and multi-epoch follow-up observations to confirm their nature as binaries with a degenerate accreting companion. The Galah survey provides a representative and statistically well-characterized sample of Galactic RGs. This enables a robust estimate of the symbiotic to any other RG ratio, which can be extrapolated to the entire Galaxy while minimizing the systematic uncertainties associated with the modeling of the population of binaries. In the course of our analysis of the spectra collected by the Galah survey, we also identified SySts in the HighAcc regime as well as burning-type systems (some of which were previously known). These sources will be discussed in a separate paper, as the present work focused solely on SySts in a LowAcc state.

Following an initial sample of 33 objects studied in Munari et al. [108], hereafter Paper I, this Paper II identifies a further 44 symbiotic stars accreting at the lowest detectable levels (a few $L_{\odot}$). These 77 accreting-only symbiotic stars already outnumber previously known cases, and the investigation of an additional large sample is well underway for the forthcoming Paper III. Recent years have witnessed a surge in the search for new SySts, as reviewed in Appendix § 11. All newly discovered SySts are almost invariably of the burning type or the HighAcc variant of the accreting-only type. To date, no other search has reached our level of sensitivity toward the LowAcc variant or provided the extensive multiwavelength follow-up observations (spanning several years) required to characterize the main properties of these new SySt systems.

Target selection

The target selection for the present study was conducted in 2019, concurrently with that for Paper I, where more comprehensive details and quantitative definitions are provided. We focus on M-type red giants, as they are the most abundant among known SySts. Systems containing other types of cool giants will be discussed in subsequent papers of this series.

Given the high galactic latitude (and consequently low reddening) of the Galah observations, red giants were initially selected by applying the following criteria: (i) 2MASS color $(J-K_{\mathrm{s}})\geq0.90$, and (ii) Gaia parallax $\varpi\leq3$ mas. To isolate those with an M-type spectrum, we inspected the Galah spectra for TiO absorption bands, in the process also picking up some S-type giants whose spectra exhibit both TiO and molecular bands associated with s-process elements (specifically ZrO). The M giants were then organized into the temperature sequence defined in Paper I, consisting of 13 $f$-bins based on the depth of the TiO band-heads. The boundaries of these bins were defined to ensure a comparable number of objects (several thousands) in each, at least for the lower $f$-bins (hotter M giants), where the majority of the population resides. Median template spectra were subsequently derived for each $f$-bin. A calibration of the $f$-bins sequence in terms of absolute magnitudes and intrinsic colors is provided in Appendix D, where we also discuss the shortcomings of the classical M0–M10 III classification scheme when applied to high-resolution spectra and their quantitative analysis.

The subsequent step involved searching the Galah spectra of these M giants for H$\alpha$ and H$\beta$ emission signatures. To achieve this, the appropriate templates were subtracted from the continuum-normalized spectra, and the residuals were inspected for H$\alpha$ emission. The resulting candidates were then all visually verified to eliminate false positives. While Paper I focused on M giants with an H$\alpha$ emission profile exceeding a threshold of 0.5 in the subtracted spectrum, the present work extends this selection to a lower threshold of 0.1.

Not all objects exhibiting H$\alpha$ emission are necessarily valid candidates for accreting-only SySts. The primary sources of false positives are radial pulsators (e.g., Miras or SRa variables), which trigger Balmer line emission via shocks forming in the deep atmosphere during the pulsation cycle. We filtered out Miras and SR variables by discarding objects whose ASAS-SN light curves are modulated by signatures of radial pulsation with an amplitude $\geq0.5$ mag. As a secondary pruning criterion, we examined the relative intensities of the H$\beta$ and H$\alpha$ emission in the subtracted spectra. Since radially pulsating stars invariably show H$\beta$/H$\alpha> 1$ on subtracted spectra, we only retained systems where H$\beta$/H$\alpha\ll1$.

It is, of course, possible for a Mira to be paired with a degenerate companion in a SySt; o Cet itself, the prototype of Mira variables, is a well-known, accreting-only SySt. Indeed, Hubble (UV) and Chandra (X-rays) have spatially resolved its WD companion, the accretion disk, and the fueling stream [70]. Candidate Galah SySts containing pulsating variables will be addressed elsewhere in this series.

The $\sim100$ final candidates selected for the present Paper II were all located at $\delta\geq-27^\circ$ to ensure accessibility from the telescopes in Asiago and Varese. In 2021, we began collecting high-resolution Echelle spectra and conducting photometric flickering runs, supplemented by UV and X-ray observations obtained with the Swift satellite. After two years, we shortlisted 60 objects selected for being the most promising and/or having the most extensive follow-up data. These were monitored intensively for an additional three years (2023–2025), ultimately leading to the discovery, validation, and characterization (over a total five-year period) of the 44 new and confirmed LowAcc SySts presented in this work and listed in Table A.1. The confirmatory and follow-up observations (2021–2025) for the new SySts are described in Sects. 3–6, with the bulk of the technical specifications provided in Appendices E–H.

Radial velocities, orbits, and H$\alpha$ emission from high-resolution Echelle spectroscopy

A large number of high-resolution optical spectra of the new 44 SySts were collected between 2021 and 2025 using the Echelle spectrographs mounted on the Asiago 1.82m and Varese 0.84m telescopes. A log of these observations is provided in Table 1. In total, Echelle spectra were obtained over 544 distinct nights,

Name	date	HJD	Varese	Asiago	Galah
GaSS 2-69	2024-08-20	2460543.37		−125.5
GaSS 2-69	2025-07-18	2460875.42		−126.8
GaSS 2-69	2025-08-07	2460895.42		−125.3
GaSS 2-69	2025-09-17	2460936.40		−123.9
GaSS 2-69	2025-10-17	2460966.36		−123.4
GaSS 2-70	2015-06-06	2457180.11			107.8
GaSS 2-70	2021-08-20	2459447.45			110.8
GaSS 2-70	2024-08-08	2460531.37		107.1
GaSS 2-70	2025-08-09	2460899.42			111.7

Notes. The full table is only available electronically via CDS, a small portion is shown here for guidance on its form and content. The RVs are expressed in km s$^{-1}$.

with an average of 12.5 visits per program star. The Galah survey provided the initial spectrum for each object (with a mean date of June 2016), contributing a further 47 epochs (one per object, plus a second visit in some cases). This brings the grand total to 591 visits, or an average of 13.5 per object. Details of the telescopes, spectrographs, and observing and data reduction procedures are provided in Appendix E.

Radial velocities

The radial velocities (RVs) were measured on the single, one-dimensional, continuum-normalized version of the Echelle spectra via Fourier cross-correlation against suitable templates [137], as implemented by RV/FXCOR in IRAF. To maximize the signal-to-noise ratio (S/N) of the object spectra and to avoid regions affected by telluric lines, interstellar NaI, and possible stellar emission lines, the cross-correlation was restricted to three spectral intervals: 4900–5800, 6000–6250, and 6330–6450 Å. The list of measured RVs is provided in Table 1.

The spectra collected with the Asiago Echelle spectrograph were all cross-correlated against the same template, selected from the synthetic spectral library of Munari et al. [107]. The template was chosen for a resolving power of 20,000, with atmospheric parameters of $T_{\rm eff}$=3500 K, $\log g$=1.5, [Fe/H]=0.0, [$\alpha$/H]=0.0, and $\zeta$=2 km s$^{-1}$. This approach has already proven to be very effective in obtaining the best available RV orbit for the M3 III red giant in the symbiotic recurrent nova T CrB from long-term Asiago and SMARTS-Chiron Echelle observations [110], avoiding systematic offsets and providing RVs accurate to $\sigma$=0.27 km s$^{-1}$ over a multi-year timescale.

A different measurement approach was adopted for the spectra collected with the Varese telescope to enhance its performance to better match that of the Asiago Echelle spectrograph, which operates at a higher resolving power. The Gaia DR3 catalog of radial velocities was searched for a suitable RV template located within 2° of each program star. For each observation of a program star, the corresponding template was observed immediately before and after. The object spectrum was then cross-correlated separately against the two spectra of the template, and the results were averaged. The median difference between these two RV measurements is 0.29 km s$^{-1}$, a stability comparable to that of the Asiago data. The multi-epoch observations of the selected templates were finally mutually cross-correlated to verify their RV constancy over the 2021–2025 interval.

Finally, the Galah radial velocities were imported from Galah Data Release 4 (Buder et al. [17]).

Spectroscopic orbits

The RVs from Table 1 were analyzed together with ASAS-SN g-band photometry (Shappee et al. [126]; Hart et al. [54]) to search for orbital motion. The orbital period was searched for simultaneously in the RVs and photometric data using the phase dispersion minimization algorithm of Stellingwerf [134] and the Fourier analysis for unequally spaced data of Deeming [36]. To be accepted as a viable photometric period, the corresponding folded light curve must display the shape expected from orbital modulation (e.g., reflection effect, eclipses, Roche-lobe filling, or transit at the inferior conjunction of the hot spot) and be properly phased with respect to the folded RV curve.

For 30 of the 44 program stars, it was possible to obtain an orbital solution, as reported in Table 2; for all of these, the photometric orbital period (where detected) converges to the same value derived from the radial velocities. An example of the derived orbits is provided in Figure 1 for the case of GaSS 2-38¹

Name	P	K	$\gamma$	e	$\omega$	T$_0$
GaSS 2-34	650	1.8	−11.8	0		2461360
GaSS 2-35	380	2.7	−35.0	0		2461480
GaSS 2-38	470	1.6	−1.3	0		2461540
GaSS 2-40	570	1.2	−46.2	0		2461590
GaSS 2-44	342	2.3	−31.5	0.2	120	2461490
GaSS 2-54	335	6.3	−67.0	0.2	190	2461650
GaSS 2-56	690	5.7	−100.8	0		2461590
GaSS 2-76	412	2.4	−86.5	0		2461610
good orbits
GaSS 2-39	2000	7.6	88.9	0		2461580
GaSS 2-45	600	1.9	133.8	0		2461625
GaSS 2-51	170	3.5	−62.0	0		2461650
GaSS 2-52	840	11.0	−60.0	0		2461640
GaSS 2-55	2250	1.6	16.0	0		2460340
GaSS 2-58	335	3.0	−106.5	0		2461514
GaSS 2-60	800	3.7	77.0	0		2461934
GaSS 2-67	450	2.0	−85.5	0		2461720
GaSS 2-68	370	2.0	116.0	0		2461620
GaSS 2-69	310	2.9	−123.0	0.2	180	2461650
GaSS 2-71	600	2.3	20.5	0		2461280
GaSS 2-73	3060	1.3	−93.0	0		2460420
GaSS 2-74	116	1.7	23.5	0		2461050
GaSS 2-75	380	2.7	−7.5	0		2461620
fair orbits
GaSS 2-41	220	1.7	94.5	0		2461530
GaSS 2-43	249	2.4	−62.5	0		2461480
GaSS 2-46	720	2.0	323.5	0		2461870
GaSS 2-48	4500	1.9	−14.3	0		2461625
GaSS 2-63	400	3.0	−94.5	0		2461414
GaSS 2-64	1350	4.0	7.0	0		2461620
GaSS 2-65	550	1.5	62.0	0		2461000
GaSS 2-70	650	2.7	109.5	0		2461520

The 470-day period of GaSS 2-38 is typical for SySts (which usually range between 1 and 3 years; Mikołajewska (2003)), and the small semi-amplitude $K$=1.6 km s$^{-1}$ suggests the combination of a low orbital inclination and a low-mass companion (for comparison, a WD of 1.0 M$_{\odot}$ at $i$=90° would have caused a semi-amplitude of $K$=14.9 km s$^{-1}$ for an RG companion of 1.5 M$_{\odot}$). The amplitude of the photometric modulation of GaSS 2-38 in Figure 1 is $\Delta g$=0.04 mag; its single maximum, centered around orbital phase 0.25, suggests that its origin is the side of the RG irradiated by the hot companion. In contrast, a single maximum around phase 0.65 would point to the source being the hot spot transiting at inferior conjunction (see Figure 8 of Munari et al. [110]), while a double maximum centered at phases 0.0 and 0.5 would support an ellipsoidally distorted RG filling its Roche lobe.

The 30 orbital solutions listed in Table 2 were graded using a quality flag. Rather than relying on a strict analytical definition—which would be difficult to formulate given the complex requirements involving period convergence between photometry and RVs, light curve morphology and phasing consistent with known SySt variability types, the number of RV data points, their rms scatter, and their orbital phase coverage—this flag reflects an overall qualitative assessment of each orbit.

For those labeled "good", it is unlikely that new observations will substantially change the orbital parameters (the period in particular). For the "tentative" group, the listed orbit is the best that can be derived given the current data limitations (too few RVs and/or low-quality ASAS-SN photometry), while "fair" orbits represent an intermediate quality level.

Quantitatively, for good orbits, the $N$ available RVs yield an average rms=0.95 km s$^{-1}$ and rms $/\sqrt{N}$=0.18 km s$^{-1}$. These values degrade to 1.60 and 0.42 km s$^{-1}$ for fair orbits and show a further increase for tentative ones.

Seven of the remaining 14 program stars show only a photometric period. For GaSS 2-37, GaSS 2-59, and GaSS 2-77 (pe-

riods of 540, 580, and 650 days, respectively), the light-curve shape resembles that of a Roche-lobe-filling RG seen at low orbital inclination, or one only partially filling the lobe (small photometric amplitude). For the other four stars in this group, namely GaSS 2-42, GaSS 2-61, GaSS 2-62, and GaSS 2-66, with periods of 340, 372, 375, and 360 days, respectively, the light curves show single maxima and minima: given the proximity to the 365-day yearly alias, a photometric coverage longer than the ∼10 years currently available from ASAS-SN would be necessary to fully cover all phases and confirm the periods. Finally, for the last seven stars (GaSS 2-36, GaSS 2-47, GaSS 2-49, GaSS 2-50, GaSS 2-53, GaSS 2-57, and GaSS 2-72), the search for a suitable periodicity in both the photometric and RV data was unsuccessful.

Hα profiles from subtracted spectra

For our targets, the Hα emission is sufficiently weak that properly detecting the line, extracting a clean profile, and measuring the radiated flux requires subtracting a template spectrum of the underlying M giant. Given the similar resolution of the Asiago and Galah spectra (e.g., Vogrinčič et al. [139]), we adopted the 13 $f$-bin Galah templates defined by Munari et al. [108]. We implemented a C++ pipeline based on minimum distance estimation (MDE) to identify the best-fit $f$-bin template by minimizing the $\chi^2$ between the observed and template spectra. The detection and analysis of Hα emission profiles are restricted to the Asiago spectra; the lower resolution of the Varese observations makes it unreliable to use the same set of Galah templates, even if appropriately degraded.

The procedure begins with initial data cleaning for cosmic rays and telluric lines, followed by a preliminary wavelength alignment. The observed spectrum is then locally normalized using a moving average and compared to the templates. The MDE algorithm iteratively optimizes both the radial-velocity shift and the Gaussian broadening required to match the instrumental resolution. During this process, the cores of the emission lines (e.g., 6561.3–6564.3 Å for Hα) are masked. After fine-tuning the normalization and resolution, the optimal $f$-bin template was identified. The S/N and instrumental stability are high enough that the optimal $f$-bin template remains constant for all the observations of a given object. Finally, the template was subtracted, leaving a pure emission residual for further analysis (the profile in blue in Figure 2).

An example of the results is presented in Figure 2 for the program star GaSS 2-35. The 15 observing epochs, spanning from 2015 to 2025, highlight the variability of Hα in both profile shape and radiated flux. Similar figures illustrating the evolution of the Hα profile for the remaining program stars are available from the same web site as given in footnote 1.¹ The basic parameters extracted from these cleaned Hα profiles are summarized in Table 3, along with the results of fitting these profiles with a combination of two Gaussians (one for the emission and one for the absorption component). We also experimented with other profile types (Lorentzian, Voigt) but found no systematic statistical improvement over Gaussian fits. The Hα profile deconvolution reported in Table 3 is intended purely to provide dataset-wide inputs for exploring the potential orbital modulations in the next section; individual fits are not intended as high-precision physical models of single profiles that would warrant formal error propagation or their explicit quoting in Table 3.

Name	tel.	date	Whole profile peak	Whole profile EW (Å)	Whole profile flux	Emission flux	Emission RV wrt RG (km/s)	Emission FWHM (km/s)	Absorption flux	Absorption RV wrt RG (km/s)	Absorption FWHM (km/s)
GaSS 2-35	Galah	2015-11-11	0.463	0.590	4.0e-14	1.5e-13	+1.0	48	1.1e-13	−0.7	40
GaSS 2-35	Asiago	2021-02-22	0.338	0.557	3.7e-14	1.2e-13	+1.3	55	8.4e-14	+0.5	45
GaSS 2-35	Asiago	2021-10-17	0.520	0.913	4.7e-14	6.8e-14	+1.3	63	2.1e-14	−1.6	37
GaSS 2-35	Asiago	2021-12-20	0.526	0.817	5.0e-14	2.0e-13	−0.7	51	1.5e-13	−1.6	44
GaSS 2-35	Asiago	2022-01-16	0.420	0.757	4.6e-14	8.6e-14	+2.3	57	3.9e-14	+1.9	39
GaSS 2-35	Asiago	2022-03-23	0.443	0.717	5.2e-14	2.2e-13	−1.8	53	1.7e-13	−2.7	46
GaSS 2-35	Asiago	2022-11-04	0.624	1.166	5.4e-14	1.0e-13	+0.0	61	4.7e-14	−0.9	44
GaSS 2-35	Asiago	2023-09-30	0.562	0.961	4.9e-14	6.9e-14	−2.4	67	2.0e-14	−2.5	29
GaSS 2-35	Asiago	2023-12-21	0.542	0.883	5.4e-14	1.7e-13	−0.2	54	1.1e-13	−1.1	43
GaSS 2-35	Asiago	2024-01-28	0.459	0.738	4.5e-14	1.4e-13	+0.2	55	9.2e-14	−0.6	41
GaSS 2-35	Asiago	2024-03-24	0.305	0.533	3.9e-14	1.2e-13	+2.5	58	8.1e-14	+2.1	44

Orbital modulation of Hα emission

To spatially locate the source of Hα emission in the 44 newly discovered LowAcc SySts, we investigated the properties of the pure Hα emission profiles isolated by subtracting the red giant contribution from the observed spectra. We focused on the 22 objects with "good" or "fair" orbital solutions (Table 2). We explored the relationship between orbital phase and the Hα parameters listed in Table 3, but found no clear pattern either on an individual object basis or when phase-averaging all systems together.

Notes. The full table is only available electronically via CDS; a small portion is shown here for guidance on its form and content. "Emission" and "Absorption" refer to fitting the "Whole profile" (blue spectrum in Figure 2) with a combination of two Gaussians, one in emission and the other in absorption. The fluxes are in erg cm$^{-2}$ s$^{-1}$, and the radial velocities are given with respect to the red giant.

The lack of clear repeatability with orbital phase for individual objects suggests that stochastic accretion variability dominates over the effects of orbital geometry. Furthermore, the absence of a distinct trend in the phase-averaged ensemble suggests a combination of accretion modes, likely involving both wind-capture and Roche-lobe overflow.

Flickering observations

The flickering observations for Paper I were acquired with a 50cm telescope, featuring a 40 arcmin, well-corrected field of view operated robotically from San Pedro de Atacama, Chile. For the present Paper II, all flickering observations were carried out robotically with the Asiago 67/92cm Schmidt telescope (focal length 2.15m). The telescope is equipped with a Moravian CCD camera (KAF-16803 detector, 4096×4096 active matrix, 9 µm pixel size), covering a field of 59×59 arcmin at 0.86 arcsec/pix.

This setup offers several advantages compared to Paper I: (1) a ∼2× gain in light-collecting power, significantly benefiting the S/N and enabling the exploration of shorter flickering time-scales; (2) a field of view twice as large (perfectly corrected by a 3-element field-flattener), allowing for a broader choice of both the photometric comparison sequence and check stars; (3) tighter control over real-time focusing and guiding (fractional-pixel stability over hours) to fully exploit the excellent seeing conditions (typically FWHM=1.1 arcsec); (4) higher overall observational accuracy, reducing the dispersion on check stars to as little as 0.002 mag over one-hour runs; and, in particular, (5) the possibility of multiple visits to each target.

In Paper I, limited time allocation and unsupervised queued-robotic mode restricted observations to a single ∼70min visit at a random epoch for most of the targets. Given the erratic nature of flickering in SySts accreting at the lowest levels, and the variable photometric quality of different nights (including seeing, guiding, and Moon phase), detecting flickering in a single visit often depended on observing the right star at the right time during a perfect night. This likely accounts for the limited number of flickering detections in Paper I.

In contrast, all program stars in this paper were visited several times between 2021 and 2024 (more than 9 times for eleven of them), increasing the probability of collecting data during both optimal photometric conditions and epochs of active flickering. Furthermore, repeated observations allowed us to refine the sampling times for subsequent visits and improve the photometric comparison sequences and check star selection. In a few cases, this involved off-centering the target within the field to compensate for the low stellar density encountered at high Galactic latitudes. The only drawback of the Asiago 67/92cm telescope (located at +46° latitude) is the mechanical constraint preventing observations of targets south of declination −23°, a limitation affecting the four southernmost program stars.

All targets were searched for flickering in the Landolt $B$ band, following the same rationale as in Paper I. Although the entire optical train of the Asiago 67/92cm Schmidt telescope is UV-transparent (e.g., the corrector plate is made of UBK7 glass), we did not attempt $U$-band observations. This was due to the prohibitive paucity of field stars (an order of magnitude less than in $B$) and the significantly lower S/N, as the targets are typically ∼2 mag fainter in $U$ than in $B$. As demonstrated by Munari et al. [110] radiative modeling of T CrB during its 2005–2015 deep quiescence, $B$-band emission is dominated by the hot spot rather than the accretion disk body or the red giant. This makes the $B$ band ideal for flickering observations of distant SySts accreting at low rates, as targeted in this paper.

The flickering observations were calibrated individually from the instantaneous local photometric system to the Landolt [78] standard system. This was necessary to achieve and maintain millimag precision over runs exceeding one hour. A purely differential approach with respect to field stars would not have reached the required precision, given the instability of the atmosphere during the observations. The fractional-pixel stability of the telescope guiding during each run ensured that the symbiotic star and all comparison and check stars remained on the same pixels, minimizing flat-fielding errors and boosting internal accuracy. The observing strategy, data reduction, and selection of comparison and check stars, along with the details of the aper-

Notes. The full table is only available electronically via CDS; a small portion is shown here for guidance on its form and content. The columns list the (1) name, (2) date of observation, (3) median HJD of the observing block, (4) exposure time of each $B$-band frame, (5) number of $B$-band frames within the block, (6) median $B$-band magnitude of GaSS target during the observing block, (7) $B$-band magnitude standard deviation of GaSS target (red star in Figure 3), (8) difference (dashed red vertical line in Figure 3) of the observed and expected magnitude standard deviation (the latter being calculated based on GaSS magnitude and the fit to check stars; i.e., dashed black line in Figure 3), (9) RMSE of the fit to check stars (the breadth of the gray band in Figure 3), (10) significance of our flickering detection ($\sigma_{\mathrm{STD}}$; see Eq. (1)), and (11) notes on the quality of the data, where $a$ indicates a low number of check stars ($<5$) and $b$ a poor distribution of check stars with brightnesses above and below the SySt.

ture and PSF-fitting photometry, are discussed in detail in Appendix F.

The results of our flickering analysis are summarized in Table 4. In Figure 3, we highlight four program stars (GaSS 2-43, 2-45, 2-46, and 2-49) exhibiting some of the most prominent flickering detections, ranging from 0.0031 to 0.0057 mag (dashed red lines).

(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
Name	date	HJD	expt (sec)	$N(B)$	$B$ (mag)	$\sigma(B)$ (mag)	diff. (mag)	RMSE (mag)	$\sigma_{\mathrm{STD}}$	notes
GaSS 2-45	2022-02-25	2459635.626	120	25	15.089	0.0081	0.0013	0.0007	1.9
GaSS 2-45	2023-02-17	2459992.524	120	25	14.690	0.0101	0.0055	0.0001	70.4
GaSS 2-45	2023-11-12	2460260.596	130	17	15.233	0.0126	0.0039	0.0014	2.8
GaSS 2-46	2022-12-25	2459938.678	480	10	15.954	0.0114	0.0057	0.0002	28.7
GaSS 2-46	2023-02-17	2459992.568	480	10	16.075	0.0042	0.0003	0.0006	0.5	b
GaSS 2-47	2022-02-24	2459634.614	30	80	12.747	0.0074	0.0024	0.0003	7.7
GaSS 2-47	2022-12-27	2459940.684	30	45	12.754	0.0099	0.0045	0.0006	7.7
GaSS 2-48	2022-03-02	2459640.582	20	120	12.091	0.0075	0.0024	0.0004	6.0
GaSS 2-48	2023-01-04	2459948.679	20	35	12.029	0.0072	0.0026	0.0007	3.6
GaSS 2-49	2022-02-25	2459635.675	90	15	13.520	0.0071	0.0031	0.0002	14.3	a

To identify flickering, we calculated the standard deviation of the $B$ magnitude for each target ($\mathrm{STD}_{\mathrm{GaSS}}$) within an observing block ($\sim70$ min) and compare it to the expected value of the standard deviation ($\mathrm{STD}_{\mathrm{fit}}$) for a non-flickering star of equivalent magnitude. The latter was derived from a second-order polynomial fit (represented by the dashed black line in Figure 3) to a sample of check stars of similar red color to the program symbiotic star (within $\Delta(B-V)\lesssim0.2$ mag) and located around it. The significance of the detection is quantified by $\sigma_{\mathrm{STD}}$, which is defined as

where RMSE is the root mean squared error of the polynomial fit. $\sigma_{\mathrm{STD}}$ is equivalent to the ratio of the length of the red dashed line to the width of the grey band in Figure 3. For cases where ($\mathrm{STD}_{\mathrm{GaSS}}-\mathrm{STD}_{\mathrm{fit}})<0$, we uniformly set $\sigma_{\mathrm{STD}}=-1$ to indicate the absence of detectable flickering. This metric effectively serves as a measure of the statistical significance of our flickering detections. The qualitative coding reported in Table A.1 corresponds to $2.5\leq\sigma_{\mathrm{STD}}<3.5$, $3.5\leq\sigma_{\mathrm{STD}}\leq5.0$, and $\sigma_{\mathrm{STD}}>5.0$ for flags "y", "Y", and "Y!", respectively.

As illustrated in Sect. F.3 and the associated Fig. F.1, the majority of the check stars are themselves red giants. Their stability over the $\sim70$ min flickering runs (black points defining the gray bands of Figure 3) proves that the flickering observed in the program stars is not an intrinsic property of red giants (either RGB or AGB), but is instead directly associated with symbiotic activity.

Near-UV excess in Gaia XP spectra

Gaia DR3 provides access to XP spectra, which are low-resolution, slitless prism spectra covering the 3,300 to 10,500 Å range in two arms—$B_P$ and $R_P$—with minimal overlap at H$\alpha$. The XP spectra are provided in a flux-calibrated format as the mean over all individual spectra collected during the first 34 months of satellite operations (2014–2017). Technical details are summarized in Appendix G.

As an example of their diagnostic potential, in the left panel of Figure 4 we compare the XP spectra of four symbiotic stars with those representative of their red giant components (see Fig. G.1). The four selected symbiotic stars exemplify those well: (i) in stable H-burning (V919 Sgr); (ii) during high accretion states (e.g., T CrB in 2015–2017); (iii) accreting at a moderate rate (AT 2023txn); and (iv) accreting at low or very low levels (SU Lyn). Note that the apparent noise shortward of 4000 Å is not intrinsic to the source but is a manifestation of the Gaia DR3 choice to represent spectra using Hermite functions (see Appendix G for a discussion).

As clear from Figure 4, a large near-UV excess characterizes both burning and high-accretion types; in the latter, the excess emission extends to redder wavelengths, while a much stronger H$\alpha$ line is typical of the former. No significant H$\alpha$ emission is detectable in XP spectra for moderate and low accretion types, which differ primarily by the amount of near-UV excess emission shortward of 4000 Å. Despite the low resolving power, Gaia XP spectra are sufficient to resolve some molecular bands and provide a spectral classification of the red giant. In addition, it is possible to detect chemical anomalies such as the S$_6$I spectrum for the red giant in SU Lyn (note the ZrO 9330 Å band, which is typical of S-type stars; see Pearse & Gaydon [117]; Wright et al. [142], a star traditionally considered a normal M6 III [101]. Simultaneous with the spectral typing of the red giant, the reddening ($E_{B-V}$) can also be well constrained thanks to the wide wavelength coverage of the XP spectra.

For the low accretion rates of interest in this series of papers, H$\alpha$ emission is too weak to provide a detectable signature in XP spectra. Consequently, the most viable indicator of an interacting binary nature is the presence of excess flux at near-UV wavelengths. To detect such an excess, it is necessary to compare the reddening-corrected spectra of the program stars with suitable templates. To build templates for all $f$-bins, we have median-combined the XP spectra of 32,182 M giants ("32k sample" hereafter) that we identified among the one million stars observed by Galah so far. We limited the selection to the 32k stars with viable XP spectra and with a reddening of $E_{B-V}<0.1$ mag as derived from Schlegel et al. [124] maps (see details in Appendix § 15).

The spectra of the newly discovered SySts were iteratively fitted to such $f$-bin templates while simultaneously adjusting the reddening correction following the extinction law of Fitzpatrick [47]. The reddening values returned by the fitting procedure align well with those derived from the Schlegel et al. [124] maps; this consistency enhances confidence in the methodology and suggests that any circumstellar reddening, if present, is modest. The fitting was restricted to $\lambda\ge4200$ Å to avoid circularity with the search for flux excess at $\lambda\le4000$ Å. This procedure yielded excellent matches, examples of which are provided in the right panel of Figure 4.

The presence of a UV excess was then evaluated by comparing the XP spectra of the matching template and the program star for $\lambda\le4000$ Å, grading it into four classes: "y" pertains to stars for which data points shortward of 4000 Å show a mean excess between 3 and 5$\sigma$ with respect to the template spread; class "Y" for 5 to 7$\sigma$; and "Y!" for excesses larger than 7$\sigma$. Examples for each class, including the case of no excess ("n"), are plotted in the right panel of Figure 4. The final results, including the matching $f$-bin, reddening, and near-UV excess classification, are summarized in Table A.1.

Swift X-ray and UV observations

All program stars were observed in X-rays and in the UV with the Swift satellite (Gehrels et al. [51]). Between March, 2022 and July, 2024, we observed 37 of the 44 program stars through two Guest Observer programs (IDs 1821023 and 1922036; PI: N. Masetti) during Swift cycles 18 and 19. Moreover, between February and June, 2024, five target-of-opportunity (ToO) observations of sources GaSS 2-53, GaSS 2-55, GaSS 2-64, GaSS 2-68, and GaSS 2-71 were acquired with the same satellite. For three additional objects (GaSS 2-40, GaSS 2-48, and GaSS 2-57), a number of serendipitous observations were retrieved from the ASI-Swift archive.² Details of the data reduction and analysis are provided in Appendix § 16, and the results are listed in Table 5.

Object	Date	start (UT)	expt (sec)	count rate$^{a}$	flux$^{b}$	N	expt (sec)	UVM2 (mag)	UVW2 (mag)	Remarks
GaSS 2-34	23 May 2023	01:56	950	<1.5	<1.3	3	935		16.36±0.04	$c$
GaSS 2-34	19 May 2024	15:51	2455			5	2433	19.55±0.13		$d$
GaSS 2-34	25 May 2024	02:42	993			1	979	19.53±0.19
GaSS 2-35	03 Apr 2022	08:45	584	<6.0	<5.3	1	576		17.97±0.09
GaSS 2-35	13 Jul 2022	15:20	522			1	514		18.05±0.10
GaSS 2-36	01 Apr 2022	12:17	489	<1.1	<1.0	2	454		18.44±0.12	$c$
GaSS 2-36	03 Apr 2022	04:16	4739			10	4614		18.45±0.05
GaSS 2-36	24 Jul 2023	19:54	657			1	646		18.62±0.12
GaSS 2-37	31 Mar 2022	01:26	1918	<3.4	<3.0	3	1886		18.73±0.08	$c$
GaSS 2-38	11 Apr 2023	19:21	918	<4.5	<4.0	1	912		17.44±0.07	$e$
GaSS 2-38	12 Apr 2023	14:33	542			2	532		17.19±0.07
GaSS 2-39	13 Apr 2023	15:56	2026	<3.3	<2.9	2	1992		18.15±0.07	$c$

Eight targets were also detected in the UV by the GALEX satellite (Martin et al. [86]) during its all-sky survey performed between April, 2003 and June, 2013. Their magnitudes in the near-UV (NUV) filter [12] are reported in Table 6. None of the program stars were detected by GALEX in the far-UV (FUV) band.

The nature of the accreting source

Given the low accretion luminosity (as inferred from the Hα emission; see Sect. 7.4) and the mean distance of 3.2 kpc (see Sect. 7.3), we did not expect to detect X-ray emission from our targets, except in two cases: (1) a given program star was undergoing a phase of greatly enhanced accretion at the time of the Swift observations, or (2) the accreting degenerate star is a neutron star instead of a white dwarf. The potential well of a neutron star is so deep that even very modest accretion rates (lower than those characterizing the program stars) would have resulted in a detectable hard X-ray source, as illustrated by the observations in Masetti et al. [87] and the modeling in Munari et al. [103].

None of the 44 new SySts were positively detected in X-rays. We conclude that none of them were undergoing a surge in accretion at the time of observations, and none harbor a neutron star; this confirms the rarity of neutron stars compared to white dwarfs as the degenerate accretors in SySts.

Ultraviolet signature of accretion

To ascertain the presence of excess emission in the Swift UVOT observations of the 44 new SySts in the M2 and W2 bands, it is first necessary to derive the level of emission expected from the photosphere of the cool giant alone and then compare it to the recorded data. To this end, we searched the Swift archive for serendipitous observations of field, high Galactic latitude, and non-variable red giants with accurate spectral classifications from Houk & Swift [62]. Such giants are sufficiently faint to avoid overloading the Swift detectors. We found 16 suitable red giants in the archive, which we supplemented with ToO observations of four additional objects (all of M spectral type). The results are summarized in Table H.1.

The level of photospheric emission in the UVOT M2 and W2 bands from these red giants can be conveniently expressed as a magnitude difference with respect to the optical $B$-band, which circumvents uncertainties related to distance. The (W2−$B$) and (M2−$B$) color indices computed from Table H.1 are plotted as a function of the spectral type in Figure 5, where we added the results for six further M-type giants obtained from their ultraviolet IUE spectra (blue points in Figure 5). For these six giants, we combined their IUE spectra with 3200–8000 Å spectra obtained with the Asiago 1.22m telescope, scaled to match the IUE flux over the 3200–3350 Å overlapping region. This procedure eliminates the effects of stellar variability between non-simultaneous observing epochs. The (M2−$B$) color index was then computed by direct integration of the passband profiles taken from Poole et al. [120] and Landolt [77], with flux zero-points set on Vega.

Swift M2 band

Since the M2 and $B$ bands lie well within the Wien tail of the energy distribution for red giants, it is expected that the (M2−$B$) index is similar for K- and M-type giants, as confirmed by panel a of Figure 5. For M-type giants, the mean value is ⟨(M2−$B$)⟩=7.44 with a dispersion σ=0.23; therefore, the $3\sigma$ limit for detecting excess emission in our program stars is (M2−$B$)≤6.70 mag. The results in Table 5 show that only three program stars were detected in the M2 band out of the 24 observed (a 13% success rate). GaSS 2-52 and GaSS 2-54, with (M2−$B$) values of 5.65 and 5.02 respectively, offer a robust detection of excess emission originating from the accretion process. For GaSS 2-34 we find (M2−$B$)=6.65, which is just over the $3\sigma$ limit, meaning a highly probable detection of an ultraviolet excess.

GALEX NUV band

The GALEX NUV band is centered at a similar wavelength to Swift M2. Most of the field red giants listed in Table H.1 were detected by GALEX in the NUV band, allowing us to explore the (NUV−$B$) index similarly to the (M2−$B$) considered above. The (NUV−$B$) values plotted in panel c of Figure 5 are, however, scattered all over the place, which suggests that GALEX NUV data are not very useful in defining a meaningful reference level for the photospheric emission of the red giants or for determining if any of the program stars in Table 6 show excess emission originating from the accretion process.

Swift W2 band

The low detection rate in the M2 band was expected given the low accretion luminosity and large distances. This prompted us to explore the W2 band despite the known red-leak affecting its transmission profile. The effective central wavelength of the W2 band is 2030 Å, compared to 2220 Å for M2, which means a reduction of 2x in the photospheric emission of the red giant. In addition, the M2 filter is affected by a somewhat lower peak transmission than W2, and within its passband lies the strong 2175 Å hump characterizing the interstellar extinction law.

We hypothesized that hot accretion regions, probed at W2, could provide a signal detectable above the red leak. At the time of the observations, in the literature there was no quantitative

Notes. The full table is only available electronically via CDS; a small portion is shown here for guidance on its form and content. N = number of exposures; $a$ = 0.3–10 keV count rate ($\times10^{-3}$ s$^{-1}$); $b$ = 0.3–10 keV flux ($\times10^{-13}$ erg cm$^{-2}$ s$^{-1}$); $c$ = no variability across UVW2 exposures; $d$ = tentative variability across UVM2 frames; $e$ = 2.5$\sigma$ variability between UVW2 observations on adjacent days.

Notes. The magnitudes are in the native AB system. †: Possible spurious detection due to its proximity with the edge of the instrument’s field of view.

observational assessment of the red leak specifically for red giants. Simulations provided limited guidance because the tabulated transmission in Poole et al. [120] ends well before the extended red-tail sensitivity of the detector.

As with the M2 and NUV bands above, the data in Table H.1 were used (panel $b$ of Figure 5) to define the (W2–B) value for pure photospheric emission from the red giant. The distribution appears quite tight, with a mean $\langle(\mathrm{W2}-B)_{\mathrm{std}}\rangle=3.87$ and $\sigma_{\mathrm{std}}=0.10$ for the stars of M spectral type.

In the W2 band we observed 26 of the 44 new SySts and detected all of them. The results are summarized in Table 5 and plotted in panel $d$ of Figure 5. They show $\langle(\mathrm{W2}-B)_{\mathrm{SySts}}\rangle=3.64$ with $\sigma_{\mathrm{SySts}}=0.17$.

While individual assessments are difficult, the mean value for SySts is bluer than the reference M-type giants by 0.23 mag (corresponding to $2.3\sigma_{\mathrm{std}}$), a statistically significant difference, suggesting that an ultraviolet excess produced by the accretion processes can be detected (although faintly) over the W2 red leak for the population of low-accretion SySts explored in this paper. For the sake of completeness, we note that for the M-type stars in Table H.1, the W1 filter (of effective central wavelength 2634 Å and also strongly affected by red-leak) provides $\langle(\mathrm{W1}-B)_{\mathrm{std}}\rangle=2.93$ with $\sigma_{\mathrm{std}}=0.05$.

Discussion

Distribution of spectral types

It is an established fact that the spectral types of red giants in well-known SySts are skewed toward cooler subtypes compared to the distribution of field giants (e.g., Kenyon [73]), with a peak around M5–M6 III (Mürset & Schmid [111]). This is typically attributed to the higher mass-loss rates of cooler red giants, which favor increased accretion onto the white-dwarf companions.

The $f$-bin distribution of the red giant components for the 44 new SySts discovered in this work is shown in the panel $a$ of Figure 6, where it is compared with the distribution of the full sample of 32k M-giants observed by Galah. The distribution of these new SySts is clearly shifted toward cooler $f$-bins, consistent with the trend observed in the well-known SySt population. This supports the view advocated by Munari [102] that the subdivision of symbiotic stars into burning-types and the various accreting-only states (high, moderate, and low) does not necessarily reflect fundamental differences in the underlying binary population. Instead, these categories likely represent different evolutionary states that each symbiotic star cycles through during its lifetime.

Great variability of the accretion levels

When SU Lyn—the prototype of LowAcc symbiotic stars—was discovered in 2015–2016 by Mukai et al. [101] as the counterpart of an unidentified hard X-ray source, its identification was made possible because the object was then experiencing suitably high levels of UV excess (see Figure 4) and H$\alpha$ emission (see Figure 7). We have since continuously monitored the near-UV excess and H$\alpha$ emission of SU Lyn (Munari et al., in prep.), finding that both have remained significantly below the 2015–2016 levels, as illustrated for several epochs in Figure 7. At the accretion levels displayed by SU Lyn from 2019–2023 (roughly 1/20th of the 2015–2016 values), the object would likely have escaped identification, and the presence of a population of LowAcc SySts might have remained unsuspected.

All 44 new symbiotic stars discovered and characterized in this paper exhibit a similarly large variability in their accretion levels, closely resembling the behavior of the prototype SU Lyn. As illustrated by GaSS 2-35 in Figure 2, the H$\alpha$ emission can vary dramatically over time, by a factor of $\sim20$ between 2022 Nov 4 and 2024 Aug 21. While H$\alpha$ emission is "off" most of the time for GaSS 2-40 and GaSS 2-67, mirroring the behavior of SU Lyn, it is typically "on" in GaSS 2-35 and GaSS 2-43, with the rest of the program stars distributed in between the two.

The presence and amplitude of flickering are also highly variable and, similarly to H$\alpha$ emission, support a large variability in the accretion levels experienced by the newly discovered SySts. For example, Figure 8 compares the search for flickering in GaSS 2-45 across two distinct epochs: 2022 Feb 25 and 2023 Feb 17. The observations were conducted identically on both dates (120 sec integrations for 25 data points in the $B$ band, covering a total duration of 70 min). The night quality – considering seeing, guiding, and atmospheric stability – was similar, as evidenced by the equivalent distribution of the check stars defining the distribution of field red stars (gray band in Figure 8). Nevertheless, GaSS 2-45 displayed significant flickering on 2023 Feb 17 (amplitude = 5.7 mmag), which was instead absent on 2022 Feb 25 (amplitude < 0.9 mmag).

When Swift/UVOT observations were executed as a series of separate exposures, the results for both the M2 and W2 filters suggest variability in excess of that attributable to the red giant’s photospheric emission (see the Remarks column in Table 5). This likely reflects the stochastic nature of accretion onto the degenerate companion.

In summary, LowAcc SySts are elusive, and recognizing them requires accurate, protracted, and multi-technique observations, as exemplified by GaSS 2-76. In addition to deriving a well-defined binary orbit, our 2021–2025 observations of GaSS 2-76 have revealed large-amplitude flickering and intense and highly variable H$\alpha$ emission; however, the Gaia DR3 XP observations from a decade earlier (2014–2017) show at most a feeble near-UV excess, through which the object would not have been distinguished from normal RGs.

Reddening and distances

The amount of interstellar $E_{B-V}$ reddening affecting the new symbiotic stars has been derived from the Galactic dust maps of Schlegel et al. [124]. It is listed in Table A.2 and its distribution is shown in the panel $b$ of Figure 6. The reddening is generally small—as expected given the Galah footprint, which avoids the Galactic plane—and is well-constrained. This is evidenced by the close agreement between the values estimated from the Schlegel et al. [124] maps and those obtained by fitting the Gaia XP spectra (see the $E_{B-V}$ values compared in the right panel of Figure 4). This consistency suggests that any circumstellar contribution is generally negligible, as is typical of SySts with a non-pulsating RG.

Distances to all program stars were derived by comparing the observed $R_P$ and $K_S$ magnitudes to the absolute magnitudes listed in Table D.1. We applied reddening corrections following the Fitzpatrick [47] extinction law, assuming the standard case of $R_V=3.1$. The resulting distance distribution is shown in the panel $c$ of Figure 6, where it is compared with distances derived from the direct inversion of Gaia DR3 parallaxes (for cases where the formal parallax error is <20%). The two distributions are similar, with mean values of 3.2 and 3.9 kpc for the photometric and parallax-based distances, respectively.

Accretion luminosities

It is common practice to relate the bolometric accretion luminosity to the luminosity radiated in a single emission line using power-law relations of the following form: $\log(L_{\mathrm{acc}}^{\mathrm{bol}})=\alpha_{\mathrm{line}}+\beta_{\mathrm{line}}\log(L_{\mathrm{line}})$. Most of the available relations focus on the H$\alpha$ emission line and were calibrated using accretion disks of protostars observed over a wide range of stellar masses (e.g., Muzerolle et al. [112]; Herczeg & Hillenbrand [57]; Mendigutía et al. [90]; Alcalá et al. [5]; Fiorellino et al. [45]). An average of these relations (in solar units) is given by

which is very close to the relation by Mendigutía et al. [90] adopted in Paper I.

Very few bolometric luminosities are available in the literature for accreting-only symbiotic stars, and these generally lack a simultaneous determination of the H$\alpha$ luminosity. Consequently, a direct recalibration of Eq. (2) specifically for SySts is currently not feasible. Instead, a more readily accessible quantity is the accretion luminosity radiated longward of 2000 Å ($L_{\mathrm{acc}}^{\lambda>2000\ \text{Å}}$), which is obtained through coordinated $UBVRI$ photometry, optical spectrophotometry (that includes H$\alpha$), and ultraviolet observations in the W2, M2, and W1 bands with the Swift satellite. By combining our data with literature values for a dozen measurements of accreting-only SySts, we derived the following preliminary relation (which we plan to investigate thoroughly in a future study):

This relation confirms the slope of Eq. (2) and is consistent with its intercept, considering that the energy radiated below 2000 Å does not contribute to $L_{\mathrm{acc}}^{\lambda>2000\ \text{\AA}}$. This close agreement suggests that, in continuity with Paper I, it is reasonable to use Eq. (2) to estimate the bolometric luminosity of the program stars in this work, even though Eq. (2) was formally calibrated on protostellar accretion disks.

The epoch Hα fluxes listed in Table 3 were used to compute the mean and peak values reported in Table A.2. These values were then converted, using Eq. (2), into the mean and peak accretion luminosities listed in Table A.2. Finally, the distribution of the 44 new SySts as a function of their mean accretion luminosity is shown in panel d of Figure 6.

The mean value of the accretion luminosity for the new SySts is confined to 0.25–2.4 $L_\odot$ (5%–95% range), a value similar to the amount radiated by T CrB during deep quiescence (see [110]), and which also compares well with the range displayed by SU Lyn, the prototype of LowAcc SySts. In fact, the accretion luminosity derived by Mukai et al. [101] for SU Lyn from 2015–2016 was ∼10 $L_\odot$, at a time when the Hα emission was higher than ever observed in any of the new SySts of this paper (see Figure 7), while a minimum value of ∼0.5 $L_\odot$ was probably reached by SU Lyn from 2019–2023, when the intensity of Hα dropped to ≈1/20 of that in 2015–2016 (see Figure 7).

The mean accretion luminosity for the newly identified SySts is lower than that of the objects in Paper I, which were characterized by either stronger Hα emission or X-ray detections. In Table A.2 we also quote the flux upper limits that have been computed from the X-ray count rates in the 0.3–10 keV band reported in Table 5 using the WebPIMMS³ online tool and assuming the X-ray spectrum of SU Lyn as in Mukai et al. [101], i.e., an absorbed thermal plasma with a temperature of $kT=17$ keV and hydrogen column density of $N_{\mathrm{H}}=2.9\times10^{22}$ atoms cm$^{-2}$. The corresponding luminosity limits were calculated according to the distances reported in Table A.2. These limits (expressed in the above table in solar luminosity units) refer to absorbed values for the X-ray fluxes; the unabsorbed ones are ∼50% larger. These upper limits to $L_{\mathrm{X}}$ are fully consistent with accretion luminosity derived from the flux in Hα.

Comparing the orbits to well-known symbiotic stars

The largest and most homogeneous set of spectroscopic orbits for well-known SySts is undoubtedly that collected at infrared wavelengths over twenty years by Fekel et al. [37, 38, 43, 42, 39, 40, 41] and Hinkle et al. [59, 58], hereafter referred to as the Fekel & Hinkle sample. This sample comprises 30 objects selected from the brightest SySts near the Sun, generally at low Galactic latitudes and therefore lying close to the Galactic plane. In contrast, our sample of 30 spectroscopic orbits in Table 2 pertains to more distant objects at higher Galactic latitudes, and consequently, greater distances from the Galactic plane.

The distributions of orbital period and semi-amplitude for our sample and the Fekel & Hinkle sample are compared in

ref-3 | webpage | | | WebPIMMS | | | | | | https://heasarc.gsfc.nasa.gov/cgi-bin/Tools/w3pimms/w3pimms.pl raw: https://heasarc.gsfc.nasa.gov/cgi-bin/Tools/w3pimms/w3pimms.pl

Figure 9. Our sample peaks at smaller semi-amplitudes and shorter orbital periods. Assuming a similar random distribution of orbital inclinations, this discrepancy could suggest either more massive RGs or lighter WDs in our sample, or a combination of both. There are, however, important aspects of sample selection and observational techniques that could artificially inflate such a discrepancy.

Our program stars were set from the beginning and the observing campaign lasted for a maximum of five years for the most observed ones, whereas the Fekel & Hinkle sample was probably drawn from a larger set of objects observed over 20 yrs. This naturally favors the predominance of shorter orbital periods in our sample compared to that of Fekel & Hinkle. It is also possible that only the systems yielding the cleanest orbits were published in the Fekel & Hinkle sample, which naturally favors larger semi-amplitudes.

There is an additional possible explanation for the larger semi-amplitudes of the Fekel & Hinkle sample, linked to the adopted observational technique. Radial velocities for the Fekel & Hinkle sample were derived from molecular CO absorption lines within a narrow interval of the infrared $H$ band (Fekel et al. [37]). The Fekel & Hinkle sample is dominated by burning and HighAcc systems, while ours is entirely composed of LowAcc objects. The companion in burning and HighAcc SySts is very hot and luminous, strongly irradiating the facing side of the red giant. It is plausible that CO is (partially) dissociated by irradiation on the side of the RG facing the hot WD. This would shift the CO absorption photocenter toward the shadowed side of the RG (away from the WD), thereby inflating the orbital semi-amplitude. This irradiation-induced off-centering might also explain another discrepancy: while all but three of our 30 orbits are circular, the majority of the 30 Fekel & Hinkle orbits are eccentric.

In conclusion, given these selection and observational biases, the extent to which the orbits of distant Galah and nearby well-known SySts are intrinsically different remains unclear. A better characterization of the Fekel & Hinkle sample’s selection function and detailed radiative modeling of irradiation effects on RVs, combined with the increased Galah sample size expected in the next paper of this series, will be essential to resolving this question.

Conclusions

Following the discovery of SU Lyn, Mukai et al. [101] predicted the existence of a previously unsuspected and potentially large population of SySts accreting at much lower levels than those known at the time. This series of papers was initiated to test this prediction. We began with high-resolution spectra collected for $\sim0.55$ million stars by the Galah survey up to 2019 (De Silva et al. [35]; Kos et al. [75]; Buder et al. [17]), from which we identified $\sim15$k M-type red giants. By subtracting median template spectra, we isolated non-pulsating M-type red giants exhibiting H$\alpha$ emission. We then subjected a sample of these candidates to extensive ground-based and space-borne observations to determine if they hosted accreting degenerate companions.

In Paper I (Munari et al. [108]), we discovered 33 new LowAcc-type symbiotic stars characterized by accretion luminosities of $L_{\mathrm{acc}}\lesssim10\,L_{\odot}$ that are similar to SU Lyn. In the present work, we shifted our search toward even lower accretion luminosities ($\leq3\,L_{\odot}$) and successfully discovered, validated, and characterized 44 new symbiotic stars out to a distance of 5 kpc. This sample of 77 new SySts significantly exceeds the total number of previously known, accreting-only Galactic symbiotic stars and constitutes the bulk of the LowAcc variety discovered to date. These objects are of particular interest because it has been argued that symbiotic binaries spend most of their lifetime in a LowAcc state (Munari [102]). Properly estimating the total Galactic SySt population requires accessing the systems currently in these faint phases.

SySts in LowAcc phases are particularly elusive; their UV excess and X-ray emission typically fall well below the detection thresholds of current satellites, except for systems very close to the Sun (e.g., SU Lyn at 0.6 kpc). Their discovery and characterization therefore rely on optical observations, which must be executed with high precision because the red giant luminosity outshines the feeble accretion luminosity by a factor of $\sim1000$.

Between 2021 and 2025, we collected an extensive dataset—spanning over 544 distinct nights—of Echelle spectra using the Asiago 1.82m and Varese 0.84m telescopes. Combined with the 47 spectra recorded by Galah from 2014–2017, this provides an average of 13.5 visits per each new SySt over a ten-year baseline. These high-resolution spectra allowed us to investigate the time variability of the H$\alpha$ flux and line profiles in detail and yielded reliable spectroscopic orbits for 22 of the 44 objects, with preliminary orbits for an additional eight. An analysis of ASAS-SN photometric data (2013–2025) revealed light curves modulated by binary effects, such as irradiation or Roche-lobe filling. The Gaia XP spectra allowed us to derive accurate $E_{B-V}$ reddening values, check for chemical peculiarities, and detect the presence of near-UV excess emission associated with accretion processes. Furthermore, high-precision $B$ and $V$ photometry from the Asiago 67/92cm telescope, with external errors controlled at the 1 mmag level, allowed us to detect accretion-induced flickering in the majority of targets (with amplitudes up to 5.7 mmag) and track its stochastic behavior. UV and X-ray observations were obtained with the Swift satellite for all 44 new SySts. The X-ray data excluded the presence of a neutron star in all targets. In the UV, three individual stars were detected in the M2 filter, while the ensemble exhibited a $2.3\sigma$ excess in the W2 filter.

The detection thresholds for H$\alpha$ emission and flickering achieved in this study are unprecedented in the literature. These capabilities were instrumental in characterizing these 44 new SySts, which represent the largest number published in a single paper and more than 10% of the total known SySts in the Galaxy and Local Group (see Belczyński et al. [11]; Merc et al. [91]; Akras et al. [3]).

Given our precise knowledge of the target-selection function and the Galah survey footprint, we are confident that upon completion of this project we will be able to provide an accurate es timate of the total number of SySts in the Galaxy. This remains a key ingredient for evaluating their role in various astrophysical contexts, such as identifying the progenitors of Type Ia supernovae or the sources of UV excess in elliptical galaxies.

Data availability

The full versions of Tables 1, 3, 4, 5, and G.1 are only available in electronic form at the CDS via anonymous ftp to cdsarc.u-strasbg.fr (130.79.128.5) or via http://cdsweb.u-strasbg.fr/cgi-bin/qcat?J/A+A/.

Acknowledgements. We would like to express our gratitude to the Referee, Dr. Jennifer Sokoloski, for her thorough, constructive, and highly insightful report that helped us clarify several points that were not fully articulated in the original submission. This work has been in part supported by INAF 2023 Mini-Grant Program (contract C93C23008470001 to UM). NM acknowledges financial support through ASI-INAF and ’Mainstream’ agreement 2017-14-H.0 (PI: T. Belloni). Part of this work is based on archival data, software or online services provided by the Space Science Data Center - ASI. GT acknowledges financial support of the Slovenian Research Agency (research core funding No. P1-0188) and the European Space Agency (Prodex Experiment Arrangement No. 4000143450). MP and BJ acknowledge the support of the Slovenian Research Agency (research core funding No. P1-0188). Support by S. Dallaporta and F. Tabacco (ANS Collaboration) is acknowledged.

The 44 newly discovered LowAcc symbiotic stars

In the tables of this appendix, we summarize the main properties of the 44 newly discovered LowAcc symbiotic stars, all characterized by accretion at the lowest detectable levels ($L_{\mathrm{acc}} \lesssim3\ L_{\odot}$), as thoroughly investigated in this paper.

(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)
name	RA (J2000)	DEC	$V$ (mag)	$B-V$ (mag)	$f$-bin	Gaia XP E($B-V$)	$\Delta$UV	Flick.	RV orbit	$H\alpha$	ECH	A2 int
GaSS 2−34	00:45:55.42	+01:30:43.9	11.092	1.673	0	0.100	y	Y!	Y!	Y!	23	001
GaSS 2−35	03:58:21.17	+22:48:46.1	12.593	1.975	3	0.450	Y!	y	Y!	Y!	52	006
GaSS 2−36	04:28:37.67	+01:54:15.0	12.849	1.809	6.5	0.142	y:	Y!	Y!		19	010
GaSS 2−37	04:34:23.46	−07:32:36.0	13.348	1.631	0	0.075	n	Y!	(P!)	Y!	12	011
GaSS 2−38	05:00:51.43	+16:07:26.6	11.857	1.911	5	0.328	Y!	Y!	Y!	y	39	023
GaSS 2−39	05:08:03.56	+16:37:46.0	12.644	1.829	7.5	0.230	Y	Y	Y!		16	025
GaSS 2−40	05:16:22.01	+19:37:00.3	11.828	2.091	6.5	0.431	Y!	Y!	Y!	y	28	028
GaSS 2−41	05:26:25.63	+05:18:47.1	12.582	1.840	4.5	0.130	n	y	y	Y!	15	030
GaSS 2−42	08:13:26.66	−04:34:22.3	11.125	1.700	8	0.034	y	y	(p)	Y!	35	045
GaSS 2−43	08:27:02.89	+10:36:03.3	12.669	1.777	2.5	0.100	n	Y!	y	Y!	16	052
GaSS 2−44	08:27:30.15	+13:26:27.1	12.067	1.606	7.5	0.010	ls	Y	Y!	Y!	35	053
GaSS 2−45	10:32:55.51	+12:51:30.4	13.557	1.494	nd	nd	nd	Y!	Y	Y!	7	075
GaSS 2−46	11:11:51.29	−00:20:59.5	14.037	1.851	nd	nd	nd	Y!	y	Y!	7	077
GaSS 2−47	12:08:10.14	−09:07:52.8	11.066	1.727	5.5	0.0	ls	Y!	Y!		20	081
GaSS 2−48	13:19:53.78	−07:56:15.6	10.392	1.673	6.5	0.0	n	Y!	y	Y	16	085
GaSS 2−49	13:55:15.97	−05:33:49.6	11.801	1.703	5	0.020	ls	Y!	Y		15	091
GaSS 2−50	15:27:23.34	−06:18:39.2	12.574	1.658	7.5	0.050	n	Y!	Y!		8	103
GaSS 2−51	15:33:14.40	−16:48:53.2	10.760	1.627	8.5	0.130	Y	nd	Y	y	12	106
GaSS 2−52	15:46:20.39	−20:07:30.9	12.283	1.662	8.5	0.043	n	y	Y	nd	10	122
GaSS 2−53	16:12:59.53	−19:17:23.6	13.963	1.859	3.5	0.390	Y	nd	nd		2	160
GaSS 2−54	16:14:15.84	−27:45:19.8	$10.666^{b}$	$2.075^{b}$	4.5	0.365	Y!	nd	Y!	nd	10	164
GaSS 2−55	16:21:39.96	−16:35:53.7	12.674	1.949	0	0.430	Y	nd	Y		5	171
GaSS 2−56	16:26:12.65	+06:23:20.3	12.801	1.727	0	0.270	Y!	Y	Y!	Y!	8	178
GaSS 2−57	16:26:16.23	−27:35:34.3	$12.881^{b}$	$2.071^{b}$	4.5	0.470	Y!	nd	nd		3	179
GaSS 2−58	16:27:54.21	+06:14:41.3	12.601	1.642	8	0.050	n	Y!	Y	Y!	10	184
GaSS 2−59	16:38:12.61	−28:08:27.0	$13.042^{b}$	$1.925^{b}$	4	0.408	Y!	nd	(P!)	nd	2	206
GaSS 2−60	16:39:03.40	−19:45:13.3	11.628	1.976	8.5	0.382	Y!	y	Y	nd	5	207
GaSS 2−61	16:56:30.77	−02:40:07.5	13.433	1.982	2.5	0.300	Y	Y!	(p)	nd	4	216
GaSS 2−62	17:05:03.81	−20:39:54.9	13.305	1.850	1	0.345	Y!		(P!)	nd	4	222
GaSS 2−63	17:07:29.22	−17:52:24.4	13.493	1.937	2	0.360	y	nd	y	nd	5	224
GaSS 2−64	17:17:26.10	−17:24:51.1	11.803	1.962	3.5	0.295	Y!		y	nd	9	231
GaSS 2−65	17:36:47.58	−08:17:20.3	12.395	2.243	8	0.610	Y		y	y	9	234
GaSS 2−66	17:42:16.07	−16:01:50.6	12.993	1.849	7	0.295	Y		(P!)	nd	5	235
GaSS 2−67	17:44:09.21	+05:44:58.3	12.938	1.754	7	0.290	Y	Y!	Y	Y!	10	236
GaSS 2−68	18:11:21.40	+06:27:15.8	12.989	1.798	4.5	0.180	y	$a$	Y	Y	11	249
GaSS 2−69	18:12:14.13	+05:34:47.5	12.183	1.756	5	0.170	Y		Y	Y!	11	250
GaSS 2−70	18:12:45.59	+06:00:58.8	13.006	1.817	6.5	0.190	y		y	y	4	251
GaSS 2−71	18:44:50.54	−22:45:13.0	10.427	1.892	nd	nd	nd	Y!	y	nd	7	256
GaSS 2−72	19:01:28.50	−26:48:21.4	$12.523^{b}$	$1.913^{b}$	0	0.360	Y!	nd		nd	4	292
GaSS 2−73	19:19:53.89	−15:45:22.8	10.341	1.689	5	0.055	Y	Y	y	nd	9	323
GaSS 2−74	19:43:27.18	+02:10:01.3	11.443	1.915	11	0.280	Y!		y	Y	15	342
GaSS 2−75	20:09:45.80	+00:51:52.7	12.331	1.689	nd	nd	nd	Y!	Y!	Y!	17	354
GaSS 2−76	21:12:29.70	+05:39:42.7	12.705	1.749	7	0.069	y:	Y!	Y!	Y!	15	356
GaSS 2−77	22:13:13.16	+07:32:39.8	11.641	1.687	5	0.052	y:		(P)	Y!	21	362

Notes. The presence of confirmatory indicators – near-UV excess, flickering, orbital motion, and $H\alpha$ variability – is summarized via qualitative flags. These indicate presence and increasing prominence or confidence of the feature, following the sequence ’y’, ’Y’, and ’Y!’. Detailed numerical values and analysis are provided in the relevant tables, sections, and appendices of this paper. "nd" indicates no data available. Columns: (1) Object name (following Paper I); (2, 3) RA and Dec from the Galah DR3 catalog; (4, 5) median $V$-band magnitude and $B-V$ color from our 2021–2025 photometric monitoring (Asiago Schmidt 67/92cm, except $b$ indicating ANS Collaborator ID 0310 telescope); (6–8) results from fitting Gaia XP spectra to $f$-bin templates: best-matching $f$-bin, inferred $E(B-V)$, and detection of near-UV excess at $\lambda\le4000$ Å ("ls" stands for low signal); (9) $B$-band flickering detection (Asiago Schmidt 67/92cm; $a$ = a field star 3.6 arcsec away and 1.79 mag fainter prevents flickering measurements); (10) RV orbit determination (Asiago 1.82m and Varese 0.84m Echelle spectra); for those lacking one, ’p’, ’P’, and ’P!’ indicate increasing prominence of an orbital photometric period in ASAS-SN data; (11) $H\alpha$ flux and profile variability on Asiago 1.82m Echelle spectra; (12) number of separate nights during which high-resolution Echelle spectra were collected (Asiago 1.82m and Varese 0.84m); (13) internal identification number.

(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)
name	$f$-bin	$E_{B-V}$ (mag)	$B_P-R_P$ (mag)	$R_P$ (mag)	$K_s$ (mag)	M2 (mag)	dist (kpc)	H$\alpha$ flux mean	H$\alpha$ flux max	$L_{\mathrm{acc}}$ (L$_\odot$) mean	$L_{\mathrm{acc}}$ (L$_\odot$) max	$L_X$ (L$_\odot$) (0.3-10 keV)
GaSS 2-34	0	0.021	1.847	9.492	7.171	19.54	1.84	2.76e-14	6.3e-14	0.3	0.8	<0.014
GaSS 2-35	2	0.271	2.338	10.411	7.567		2.49	4.10e-14	5.4e-14	1.8	2.5	<0.10
GaSS 2-36	6	0.179	2.441	10.722	7.827		4.07	1.24e-14	2.1e-14	1.2	2.1	<0.052
GaSS 2-37	0	0.063	1.811	11.777	9.494		5.17	5.64e-15	1.0e-14	0.6	1.2	<0.25
GaSS 2-38	5	0.341	2.459	9.607	6.558	(>19.1)	2.02	1.50e-14	3.4e-14	0.5	1.1	<0.051
GaSS 2-39	7	0.270	2.663	10.243	7.152		3.20	5.17e-15	2.3e-14	0.3	1.7	<0.093
GaSS 2-40	7	0.492	2.742	9.263	6.035	>21.09	1.73	5.75e-15	2.2e-14	0.2	0.7	<0.0070
GaSS 2-41	5	0.190	2.206	10.660	7.921		3.86	6.46e-15	1.9e-14	0.5	1.7	<0.12
GaSS 2-42	8	0.049	2.490	8.892	6.059	(20.1)	2.21	4.85e-14	7.6e-14	1.0	1.7	<0.035
GaSS 2-43	3	0.037	2.012	10.863	8.403	(>19.1)	4.21	2.83e-14	4.8e-14	2.2	3.9	<0.22
GaSS 2-44	7	0.044	2.388	9.993	7.243	>19.97	3.53	4.28e-14	1.3e-13	2.4	8.1	<0.13
GaSS 2-45	8	0.025	2.430	11.134	8.412	>19.67	6.44	1.01e-14	1.7e-14	1.8	3.1	<0.74
GaSS 2-46	3	0.032	2.093	12.198	9.760		7.86	5.81e-15	8.4e-15	1.5	2.3	<0.68
GaSS 2-47	6	0.080	2.226	9.236	6.621	(20.4)	2.32	6.30e-14	9.7e-14	1.6	2.6	<0.30
GaSS 2-48	7	0.035	2.261	8.393	5.736	>19.23	1.74	3.00e-14	9.9e-14	0.3	1.3	<0.030
GaSS 2-49	5	0.029	2.154	9.917	7.361	>19.43	3.14	2.90e-14	4.3e-14	1.2	1.8	<0.21
GaSS 2-50	8	0.089	2.438	10.178	7.363		3.91	2.28e-14	5.1e-14	1.7	4.1	<0.23
GaSS 2-51	9	0.098	2.776	8.199	5.219		1.60	6.35e-14	1.1e-13	0.8	1.4	<0.047
GaSS 2-52	9	0.129	2.579	9.908	6.933	19.6	3.47	3.20e-14		2.1		<0.26
GaSS 2-53	2	0.409	2.292	11.959	9.020	>19.43	4.57	7.10e-15		1.4		<0.24
GaSS 2-54	4	0.436	2.469	8.484	5.401	17.76	1.01	2.00e-13		2.1		<0.016
GaSS 2-55	0	0.289	2.190	10.650	7.991	>20.36	2.47	2.53e-14	3.1e-14	1.1	1.4	<0.073
GaSS 2-56	0	0.062	2.021	11.034	8.615		3.56	2.51e-14	3.4e-14	1.4	2.0	<0.26
GaSS 2-57	4	0.659	2.604	10.504	7.175		2.13	3.80e-14		3.0		<0.046
GaSS 2-58	9	0.051	2.511	10.135	7.305	>19.94	4.17	1.24e-14	3.4e-14	0.9	2.8	<0.35
GaSS 2-59	4	0.537	2.473	10.867	7.739	>19.13	2.84	1.30e-14		1.3		<0.22
GaSS 2-60	8	0.517	3.014	8.976	5.521	>19.07	1.50	1.52e-14	3.0e-14	0.4	0.8	<0.078
GaSS 2-61	3	0.249	2.295	11.512	8.684	>20.02	4.60	2.02e-15	6.6e-15	0.2	0.9	<0.32
GaSS 2-62	2	0.266	2.213	11.180	8.415		3.62	1.25e-14	1.4e-14	1.1	1.3	<1.0
GaSS 2-63	2	0.324	2.281	11.464	8.588	>19.39	3.89	5.33e-15	7.9e-15	0.6	0.9	<0.62
GaSS 2-64	4	0.217	2.283	9.734	6.952	>20.64	2.20	5.50e-14	6.2e-14	1.7	1.9	<0.050
GaSS 2-65	8	0.651	3.155	9.439	5.788	>20.19	1.64	8.57e-15	2.1e-14	0.3	0.9	<0.033
GaSS 2-66	7	0.381	2.661	10.554	7.408		3.42	1.72e-14	3.1e-14	1.9	3.5	<0.92
GaSS 2-67	9	0.184	2.684	10.478	7.448	>20.09	4.31	3.46e-15	1.5e-14	0.3	1.7	<0.26
GaSS 2-68	5	0.207	2.244	10.922	8.123	>19.96	4.25	4.01e-15	1.0e-14	0.4	1.1	<0.20
GaSS 2-69	5	0.176	2.282	10.135	7.336	>19.1	3.01	1.69e-14	2.8e-14	0.9	1.5	<0.17
GaSS 2-70	7	0.187	2.478	10.737	7.866	(>19.1)	4.44	3.73e-15	6.3e-15	0.4	0.7	<0.34
GaSS 2-71	5	0.371	2.517	8.032	4.935	>20.66	0.95	7.70e-14	1.1e-13	0.6	0.8	<0.010
GaSS 2-72	0	0.231	2.110	10.759	8.178	>19.29	2.73	2.70e-14		1.3		<0.26
GaSS 2-73	5	0.136	2.148	8.443	5.836	>20.23	1.48	6.88e-14	1.2e-13	0.8	1.4	<0.021
GaSS 2-74	10	0.303	3.341	8.285	4.821	>20.62	1.42	1.97e-14	4.3e-14	0.3	0.6	<0.024
GaSS 2-75	9	0.103	2.655	9.726	6.765		3.25	3.33e-14	6.9e-14	1.8	3.9	<0.16
GaSS 2-76	6	0.094	2.340	10.605	7.843	>19.58	4.18	9.96e-15	1.9e-14	0.8	1.6	<0.38
GaSS 2-77	5	0.052	2.152	9.740	7.128	(20.5)	2.82	1.21e-14	3.6e-14	0.4	1.3	<0.073

Notes. Columns: (2) $f$-bin spectral classification from the depth of TiO bands as measured on Galah spectra, following the numerical expressions in Paper I; (3) interstellar $E_{B-V}$ reddening from the Galactic dust maps of Schlegel et al. [124]; (4) and (5) Gaia DR3 ($B_P-R_P$) color index and $R_P$ magnitude; (6) 2MASS $K_s$ magnitude; (7) UVOT M2 magnitude as measured by Swift, summarised from epoch observations listed in Table 5, with number in parenthesis providing GALEX NUV values from Table 6; (8) distance computed from $E_{B-V}$, $R_P$, and $K_s$ against the absolute magnitudes listed in Table D.1; (9) and (10) mean and max value for the integrated flux (in erg cm$^{-2}$ s$^{-1}$) of H$\alpha$ emission line from epoch values in Table 3 (when only one measurement is available, the ’max’ column remains empty); (11) and (12) accretion luminosity (in solar units) inferred from the H$\alpha$ flux following Eq. (2); (13) absorbed luminosity in the X-rays over 0.3-10 keV interval from the Swift XRT epoch observations listed in Table 5 (see Sect. 7.4).

Object	NUV mag.	Object	NUV mag.
GaSS 2-34	20.92±0.23	GaSS 2-47	21.23±0.28
GaSS 2-39†	22.39±0.50	GaSS 2-48	20.64±0.13
GaSS 2-42	21.82±0.31	GaSS 2-74	22.46±0.39
GaSS 2-44	22.45±0.34	GaSS 2-77	22.15±0.34

	$L_{\mathrm{acc}} = 1\,L_{\odot}$	$L_{\mathrm{acc}} = 10\,L_{\odot}$
$M_{\mathrm{WD}} = 0.6\,M_{\odot}$	90	9
$M_{\mathrm{WD}} = 1.0\,M_{\odot}$	1475	148
$M_{\mathrm{WD}} = 1.3\,M_{\odot}$	6235	624

Lifetime in the accreting-only phase

By using standard relations, we can estimate the order of magnitude for the time a symbiotic star spends accreting versus burning. We assume the accreting WD can begin burning material on its surface without a mass-expelling flash (as in V4368 Sgr or V1016 Cyg, cf. [102 figure 6.5 there]), and that the amount of mass transfered during the burning phase is sufficiently small to be ignored.

Indicating with $\Delta M_{\mathrm{crit}}$ the amount of matter a WD has to accrete to reach ignition conditions for nuclear burning at the surface, the time it takes to exhaust the burning is $t_{\mathrm{burn}} = \Delta M_{\mathrm{crit}}/\dot{M}_{\mathrm{nucl}}$, where $\dot{M}_{\mathrm{nucl}}$ is the mass burning rate. The time required to accrete the shell is $t_{\mathrm{acc}} = \Delta M_{\mathrm{crit}}/\dot{M}_{\mathrm{acc}}$, where $\dot{M}_{\mathrm{acc}}$ is the mass accretion rate. Therefore, the ratio of the time spent accreting the shell to the time spent burning it is:

This ratio also represents the expected proportion of SySts in accreting-only conditions relative to burning systems.

For stable burning conditions, the mass burning rate is governed by the core mass-luminosity relation, which – following Paczyński [116] – can be expressed as:

Considering that burning the hydrogen content of 1 gr of solar chemical mixture generates $4.6 \times10^{18}$ erg, the mass burning rates for WDs of 0.6, 1.0, and 1.3 $M_{\odot}$ are $6.1 \times10^{-8}$, $3.75 \times10^{-7}$, and $6.1 \times10^{-7}$ $M_{\odot}\,\mathrm{yr}^{-1}$, respectively.

The mass accretion rate can be expressed as:

and the standard mass-radius relation for WDs is provided by Eggleton’s approximation ([138]):

where $M_{\mathrm{Ch}} \approx1.44\,M_{\odot}$ is the Chandrasekhar limit. The radii for WDs of 0.6, 1.0, and 1.3 $M_{\odot}$ are 0.013, 0.008, and 0.004 $R_{\odot}$, respectively.

Accretion luminosities $L_{\mathrm{acc}} = 1$ and $10\,L_{\odot}$ may be considered representative of SU Lyn and the new symbiotic stars studied in the present paper and in Paper I. From the expressions above, $L_{\mathrm{acc}} = 1\,L_{\odot}$ corresponds to mass accretion rates of $6.88 \times10^{-10}$, $2.54 \times10^{-10}$, and $9.78 \times10^{-11}$ $M_{\odot}\,\mathrm{yr}^{-1}$ for WDs of 0.6, 1.0, and 1.3 $M_{\odot}$, respectively; these rates are a factor of 10 higher for $L_{\mathrm{acc}} = 10\,L_{\odot}$. The corresponding $t_{\mathrm{acc}}/t_{\mathrm{burn}}$ ratios are summarized in Table B.1: it is evident that the burning phase is much shorter than the accretion phase for the LowAcc SySts discovered in this series of papers.

The values in Table B.1 assume that the accretion phase progresses at a constant rate for the entire duration of $t_{\mathrm{acc}}$. For systems with an accretion disk, this implies that the mass transfer rate to the disk is constant over time and equal to the mass transfer rate from the disk to the central WD. This may not be always the case, though. In T CrB (in which a well structured accretion disk is fed by Roche-lobe overflow), the accretion phase does not proceed at a constant rate but cycles through LowAcc and HighAcc states ([80, 83, 110]). The LowAcc states appear to correspond to the gradual buildup of mass in the accretion disk, with little material actually transferred to the central WD. Only when the disk grows beyond stable conditions does it flush most of its content onto the WD in a brighter and shorter-lived HighAcc phase. Such a dual accretion regime effectively increases the time-averaged $\langle L_{\mathrm{acc}}\rangle$, thereby shortening $t_{\mathrm{acc}}$ and reducing the $t_{\mathrm{acc}}/t_{\mathrm{burn}}$ values reported in Table B.1.

The magnitude of this reduction remains unclear, as it depends on the buffering properties of the accretion disk. These properties are generally unknown for the conditions prevailing in SySts, particularly for WDs accreting via wind-capture rather than Roche-lobe overflow.

$f$-bin	Num.	$M(V)$	$M(RP)$	$M(K_s)$	$M(W2)$	$B-V$	$B_P-R_P$	$J-K_s$	$H-K_s$	$W1-W2$	$W2-W3$	$W2-W4$
0	984	−0.591	−1.947	−4.078	−4.02	1.511	1.661	0.939	0.192	−0.14	0.15	0.27
1	710	−0.672	−2.089	−4.279	−4.22	1.536	1.716	0.963	0.204	−0.15	0.15	0.25
2	594	−0.721	−2.215	−4.462	−4.40	1.563	1.778	0.982	0.213	−0.16	0.15	0.25
3	478	−0.834	−2.381	−4.695	−4.63	1.585	1.846	1.008	0.221	−0.16	0.16	0.27
4	423	−0.873	−2.463	−4.839	−4.78	1.609	1.919	1.028	0.230	−0.15	0.16	0.28
5	396	−0.944	−2.656	−5.117	−5.07	1.616	2.005	1.053	0.243	−0.14	0.16	0.28
6	341	−0.963	−2.735	−5.272	−5.20	1.626	2.095	1.069	0.254	−0.13	0.17	0.29
7	283	−0.958	−2.878	−5.481	−5.44	1.628	2.215	1.094	0.262	−0.12	0.17	0.30
8	254	−0.887	−2.957	−5.654	−5.64	1.623	2.352	1.114	0.276	−0.10	0.17	0.31
9	238	−0.766	−3.035	−5.859	−5.85	1.599	2.556	1.138	0.294	−0.08	0.17	0.32
10	190	−0.569	−3.101	−6.106	−6.13	1.582	2.840	1.173	0.313	−0.02	0.19	0.33
11	96	−0.309	−3.166	−6.376	−6.47	1.570	3.168	1.202	0.335	0.04	0.21	0.35
12	50	−0.005	−3.103	−6.468	−6.61	1.561	3.348	1.227	0.352	0.06	0.27	0.38

Notes. $B$ and $V$ bands are on the Landolt [78] system, as realized all-sky by the APASS survey, $B_P$ and $R_P$ are from Gaia DR3, $JHK_s$ from 2MASS, and $W1$–$W4$ from AllWISE (the latter quoted to two decimal places due to lower intrinsic photometric accuracy). The error on the quoted values is approximately 0.001 mag for APASS, Gaia, and 2MASS data (except for $f$-bin=11, where it is 0.003 mag, rising to 0.005 mag for $f$-bin=12) and $<0.01$ mag for AllWISE data (0.02 mag for $f$-bin=12).

Recent searches for new symbiotic stars

In recent years, substantial effort has been invested in discovering new symbiotic stars, employing a wide range of approaches.

In addition to a number of more-or-less serendipitous discoveries of individual new SySts (e.g., Lancaster et al. [76]; Merc et al. [94, 92, 93]; Munari et al. [109, 103, 104]; Maslennikova et al. [89, 88]; Yu et al. 2022; Petit et al. 2023; Tatarnikov et al. [136]; Bond et al. [14]), the majority of new identifications in recent years has been provided by targeted efforts digging or cross-matching existing stellar catalogs, such as:

• 16 new SySts were identified by Corradi et al. [27, 30, 29] and Rodríguez-Flores et al. [121] in the IPHAS H$\alpha$, $r$, $i$ northern Galactic plane survey and confirmed via single-visit, low-res spectra (hereafter SiVi-LowRes);

• investigation of sources listed in the SuperCOSMOS H$\alpha$ survey led Mizalski et al. (2013), Mizalski & Mikołajewska (2014), and Merc et al. [95] to discover and SiVi-LowRes confirm 20, 12, and 14 new SySts, respectively;

• cross-matching of 2MASS, WISE, IPHAS, VPHAS+, and GALEX catalogs led Akras et al. [2] and Akras [1] to discover and SiVi-LowRes confirm additional 7 new SySts;

• by working on Galah spectra of Galactic red giants, 33 new SySts were identified and confirmed by follow-up high-resolution, X-ray/UV, and flickering observations by Munari et al. [108];

• by digging the LAMOST spectral database, Jia et al. [65], Zhao et al. [148], and Chen et al. [23] identified a total of 9 new SySts;

• Botello et al. [15] discovered and SiVi-LowRes confirmed two new SySts among IPHAS and UVEX footprint;

• Lucy et al. [82] investigated outliers in the SkyMapper photometric catalog and their rapid variability among the three 20-min consecutive exposures collected for the $uvg$ bands, and found 12 new SySts confirmed by SiVi-LowRes, X-rays and flickering observations.

In addition to these validated new SySts, most of the above papers also proposed a number of candidate SySts for later scrutiny, while other papers limited to compile lists of possible new SySts without subjecting them to confirmatory follow-up observations, like the 72 objects isolated by Akras et al. [4] from 2MASS and WISE catalogs, the 25 ones found by Ball et al. [10] in the cross-match of Gaia DR3 products with X-rays and infrared surveys, the 10 sources found by Xu et al. (2024) via a joint analysis of GALEX, XMM, and Gaia DR3 catalogs, or the 12 objects proposed by Contreras Rojas et al. [26] from combined analysis of Gaia, 2MASS, WISE, and SkyMapper data.

Absolute magnitudes and intrinsic colors for $f$-bin spectral standards

The definition of the 13 $f$-bins introduced in Paper I – based on the depth of TiO band-heads measured from Galah high-resolution spectra – was adopted to overcome the inherent limitations of the classical M0–M10 III classification system. While widely utilized, the M0–M10 III sequence lacks the sharp, unambiguous definitions required for high-resolution spectroscopy and suffers from a lack of portability to red wavelengths.

In fact, the original MKK system (comprising the M0–M10 III sequence) was established using low-resolution, blue-range (3800–4900 Å), prismatic spectra, where the dispersion is strongly dependent on wavelength. This system was inherently dependent on the specific dispersion used (quoted as Å/mm at $H\gamma$), as it was not based on numerical classification criteria (such as intensity ratios of selected absorption lines or molecular bands). Instead, it required "a predefined set of standard stars to be reobserved with the given instrumentation and look for the best overall match with the spectrum of the star under investigation" [100]. The concept of "best overall match" remained unchanged in later refinements [e.g. 72] and during the transition to digital or grating spectra at longer wavelengths [e.g. 64], with the definition of diagnostic line or band ratios remaining qualitative rather than quantitative [e.g. 144].

To further complicate the application of the M0–M10 III sequence to Galah spectra, several conflicting versions of the sequence exist that are mutually incompatible [e.g. 64, 48]. For instance, an M2 III type in the Mt. Wilson system [66, 96] corresponds to M0.6 III in the Case Western Reserve version [113, 114]. Such discrepancies have significant ramifications for the application to all-sky spectral surveys and homogeneity across both hemispheres [e.g. 60, 61], to spectral atlases [e.g. 125, 144], and the calibration of absolute magnitudes and intrinsic colors [e.g. 49, 131, 135].

Calibrating the 13 $f$-bins in terms of absolute magnitudes and reddening-free intrinsic color indices is therefore of primary interest. To this end, we selected M-type giants from the Galah 32k sample (see Sect. 5) that satisfy three criteria: a Gaia DR3 parallax error of less than 10%, an interstellar reddening upper limit of $E_{B-V}\leq0.1$ mag [124], and absence of $H\alpha$ emission even at the weakest detectable level on the template-subtracted spectra. This calibration was not feasible in Paper I because the initial M-giant sample was smaller (15k versus the current 32k) and, crucially, Gaia DR2 parallaxes were significantly less accurate than the current DR3 values for distant objects.

A total of 5,040 M-giants satisfied these parallax and reddening cuts. Their distribution across the $f$-bins is reported in Table D.1, which lists the mean absolute magnitudes and intrinsic colors computed from the 2MASS [31], APASS DR8 [55, 56], and AllWISE [32] catalogs, in addition to Gaia $B_P$ and $R_P$. Reddening corrections for 2MASS $JH_K$ and APASS DR8 $BV$ data were applied using the temperature-dependent relations from Fiorucci & Munari (2003). Corrections for Gaia $B_P$ and $R_P$ followed the polynomial expressions from Gaia GSP-Phot [7], while for the AllWISE W1, W2, W3, and W4 bands, we adopted the temperature- and extinction-dependent coefficients from Zhang & Yuan (2023). The distributions around the mean values for the 5,040 M-giants are remarkably sharp and Gaussian, showing no significant outliers, as illustrated by the sample distributions in Fig. D.1. The progression of absolute magnitudes and intrinsic colors along the $f$-bin sequence is smooth and gradual as shown by Fig. D.2.

High-resolution Echelle spectroscopy

The 1.82m Copernico telescope, located in Asiago (elevation 1370m, Venetian Prealps), is operated by the National Institute for Astrophysics (INAF, Italy) through the Astronomical Observatory of Padova. The 3550–7100 Å interval is covered by the REOSC Echelle spectrograph in 32 orders without inter-order gaps, using an Andor DW436-BV camera (equipped with an E2V CCD42-40 AIMO CCD, 2048×2048 pixels, 13.5 μm pixel size). The resolving power is $R\sim22{,}000$ for a 1.8-arcsec slit width. The slit is typically rotated to the parallactic angle to compensate for atmospheric chromatic dispersion when observing at low elevations.

Echelle spectra were also collected with the Varese 0.84m telescope, operated by the ANS Collaboration at the Schiaparelli Observatory in Campo dei Fiori (elevation 1225m, Varese Prealps). This facility is equipped with a Mark III Multi-Mode Spectrograph (MMS) from Astrolight Instr. The camera at the Echelle focus of the MMS is a SBIG STX10MXE CCD; the recorded 4250–8850 Å range is covered in 27 orders without inter-order gaps. A 1×1 binning and a 1.8-arcsec slit width provide a resolving power $R\sim17{,}000$, which decreases to $\sim10{,}000$ when using a 2.0-arcsec slit and 2×2 binning.

Data reduction for both telescopes was performed using IRAF following standard procedures, including bias, dark, and dome-flat field corrections. The slit height of 22 arcsec on both instruments allows for 8 arcsec of clear sky on either side of the stellar spectrum. This ensures an accurate definition and subtraction of the sky background, which is particularly important during moonlit nights given the faintness of many program stars. Wavelength calibration was achieved by obtaining Thorium-Argon lamp exposures before and after each science frame. Flux calibration was performed using spectrophotometric standards observed throughout the night, followed by heliocentric velocity correction. Finally, the individual Echelle orders were merged into a single one-dimensional spectrum, provided both as a flux-calibrated version (primarily for investigating H$\alpha$ emission) and a continuum-normalized version (adopted for radial velocity measurements). The zero-point of the fluxed version was checked (and if necessary corrected) against the near-simultaneous $g$-band photometry provided by the ASAS-SN sky patrol ([126, 54]).

Photometric observations in search of flickering

Observing mode and data reduction

A typical flickering run acquired for this paper lasted 70min, with observations in the $B$ band interspersed with those in $V$ in a $VBB\cdots BBV - VBB\cdots BBV$ pattern. The maximum number of consecutive $B$-band exposures was 10, provided the time interval between successive $V$-band exposures did not exceed 10min. The exposures in $V$ were required to build the $(B-V)$ color base to construct and solve the color equations necessary to transform each individual data point from the local instantaneous photometric system to the Landolt [78] standard one. In solving the color equations, the $V$ frame closest in time to the given $B$ frame was chosen. The color equations are as usual: $B=b+\alpha_b\times(b-v)+\gamma_b$, $V=v+\alpha_v\times(v-i)+\gamma_v$, and $B-V=\beta_{bv}\times(b-v)+\delta_{bv}$, where $b,v$ denote magnitudes in the instantaneous local system, and $B,V$ those in the Landolt standard system. The color $B-V$ is derived directly and not computed as the difference of $B$ and $V$ values.

Aperture photometry was carried out for all targets, with the aperture radius set to $1.0\times$FWHM, and the inner and outer radii of the sky annulus set at $2.5\times$ and $3.5\times$FWHM, respectively. The target GaSS 2-68 has an optical companion fainter by $\Delta B=1.79$mag at 3.65 arcsec distance, which spoils the accuracy of the aperture photometry. We measured the images of GaS$S$ 2-68 via PSF-fitting, but the limited accuracy of the method resulted in RMSE$=0.011$ (cf. Eq. (1)), too large to allow detecting any flickering with an amplitude similar to that of the other program stars.

In order to achieve a total error budget as small as 0.002–0.003 mag on the target and the check stars, the Poissonian contribution cannot exceed 0.001 mag; this requires adjusting the $B$-band exposure time to the target brightness, ranging from 10sec for the brightest targets to 480sec for the faintest ones. It is therefore clear that the presence of flickering is probed over very different time-scales depending on the $B$-band brightness of the target.

The flickering light curves secured for the present study are similar to those presented in Paper I, with its figure 10 showing illustrative examples, where the effect of the observational cadence and the interleaving of $B$- and $V$-band images can be readily appreciated.

Comparison sequences

A total error budget as small as 0.002–0.003 mag for the whole 70min run requires the definition of a well-chosen and richly populated local $B,V$ photometric sequence (LPS), as the all-sky photometric approach is unable to guarantee such protracted precision.

The LPSs for all program stars were selected from data release DR8 of the APASS all-sky $BVgri$ survey (cf. Henden & Munari [55]), and composed of at least 50 non-variable stars widely distributed in color and generally $\sim1$ mag brighter than the program star. A preliminary observation of each target checked if any of the stars selected for the LPS deviated by $\ge3\sigma$ from the linear color relations; such stars were subsequently eliminated from the LPS. Once a pruned LPS was established, it remained fixed for all visits to the given target.

Check stars and the red giants among them

To establish the presence of flickering, the dispersion $\sigma(B)$ of the measurements during the 70 min visit to a target must be compared to the $\sigma$ of suitable check stars (CKSs), which are not part of the LPS to avoid circular arguments. To attain the desired accuracy, the color of the CKSs has to be close to that of the target (i.e., very red) and they need to be well distributed in magnitude around that of the target to properly define the shape and breadth of the Poissonian relation (the gray band in Figure 3). The choice of the CKSs is made for each target by selecting those with the most suitable $B$ and $B-V$ values among field stars, as measured during the preliminary observation used to prune the LPS.

The CKS sample comprises both red dwarfs and red giants, as illustrated in Fig. F.1. Here, all CKSs with an accurate Gaia DR3 parallax (fractional error less than 20%) are plotted in an infrared version of the Hertzsprung-Russell diagram based on 2MASS $JHK_{\mathrm{s}}$ data. This choice minimizes the impact of interstellar reddening, which is inherently low given that the Galah program stars are located away from the Galactic plane. As expected from the fixed range in magnitude of Galah stars ($12\le V\le14$ mag) and the spatial distribution of Galactic stellar populations (e.g., Zwitter et al. [149]), the highest fraction of red dwarfs among the CKSs is observed at high Galactic latitudes, whereas red giants dominate at lower Galactic latitudes as clearly shown in Fig. F.1.

The fact that the giants within our CKS sample cleanly populate the entirety of the red giant branch, and yet follow the same Poissonian relation as the red dwarfs (the gray band in Figure 3), demonstrates that the flickering observed in the program stars is not an intrinsic property of red giants (either RGB or AGB), but is instead directly associated with symbiotic activity.

Template Gaia XP spectra mapping the $f$-bin classification system

During the final stages of the Gaia payload design, the originally proposed 14-band photometric system (Jordi et al. [69]) was replaced by low-resolution, slitless prism spectra. These consist of a blue arm ($B_{\mathrm{P}}$, 330–680 nm) and a red arm ($R_{\mathrm{P}}$, 640–1050 nm), with minimal overlap near H$\alpha$. The nominal resolution $R$ ranges from 20 to 100 for $B_{\mathrm{P}}$, and from 70 to 100 for $R_{\mathrm{P}}$. Rather than standard flux tables, Gaia spectra are stored as coefficients of a linear combination of Hermite functions (55 for each arm) to optimize storage and computation. For Gaia DR3, these coefficients were converted into sampled XP spectra (flux in W m$^{-2}$ nm$^{-1}$ vs. wavelength in nm), provided at 10 nm steps from 336 to 1020 nm. These mean spectra, based on the first 34 months of operations, are available for $\sim220$ million sources with $G\le17.65$ mag. Detailed descriptions and validation are provided in Carrasco et al. [20], De Angeli et al. [33], Montegriffo et al. [99] and Babusiaux et al. [9].

Two aspects of XP spectra are particularly relevant for SySt studies: (i) the slitless prism dispersion strongly compresses the data toward longer wavelengths (especially in the $B_{\mathrm{P}}$ arm), making the spectral resolution highly wavelength-dependent, and (ii) the Hermite function representation – while efficient for smooth stellar continua – struggles to reproduce sharp features such as strong emission lines or "sawtooth" molecular bands (e.g., the TiO $\alpha$-system bandheads at 4761, 4955, and 5167 Å). Consequently, Gaia XP spectra differ significantly from standard optical spectra, as shown for the M0–M8III sequence in Fig. G.1. Therefore, Gaia XP spectra of SySts must be evaluated against templates built within the same XP framework.

To construct our templates, we retrieved XP spectra for the 32k sample Galah M-giants defined in Sect 5. We retained only sources with low reddening ($E_{B-V} \le0.1$), high S/N, and no quality flags (about 500 each $f$-bin). After correcting for interstellar reddening using Schlegel et al. [124] maps and the standard $R_V=3.1$ extinction law by Fitzpatrick [47], we medium combined the spectra into mean templates for each $f$-bin. The resulting 13 template spectra are shown in Fig. G.2 and tabulated in Table G.1. They exhibit remarkably smooth transitions between $f$-bins, preserving subtle features at blue wavelengths that might otherwise be mistaken for noise upon cursory inspection. The impact of individual anomalies among the source stars is removed by the median-combination of approximately $\sim500$ individual spectra into each template. Consequently, any absorption features typical of S-type giants (whose frequency in the field is $\sim1\%$ of that of normal M-type giants) are automatically filtered out, as is the presence of emission lines (which may occur in the small fraction of input giants that are Mira pulsating variables, of the order of a few percent of the total according to Habing [53]) or the signature of companions (primarily main-sequence stars).

X-rays and ultraviolet observations with the Swift satellite

The Swift observations were acquired with the on-board instruments X-Ray Telescope (XRT; Burrows et al. [19]) and UltraViolet Optical Telescope (UVOT; Roming et al. [122]). The XRT covers the 0.3–10 keV X-ray band, whereas UVOT data were collected mainly using UVM2 or UVW2 ultraviolet filters, with a few cases acquired in the UVW1 and U bands (see Poole et al. [120]; Breeveld et al. [16], for details). The two instruments observed all objects simultaneously, with the exception of GaSS 2-51 for which no UVOT data were acquired due to instrument safety reasons because of the close proximity of the target to the 5th magnitude star $\zeta$ Lib. Exposure times for each pointing are reported in Table 5. All observations were reduced within the fTOOLs environment (Blackburn [13]); details on the data analysis process can be found in Paper I.

Targets were generally observed in ‘snapshot’ mode, i.e. with pointings of duration less than 1000 s. When needed, especially for the X-ray data, observations were stacked together to increase the signal-to-noise ratio. We remark that the choice of an aperture photometry radius larger than usual (5″) was needed for some UVOT pointings, especially those acquired starting with August 2023 in order to address the excess broadening of the UVOT images point spread function produced by issues in the spacecraft attitude control (Cenko [21]; Hussenot-Desenonges et al. [63]; Srinivasaraghavan et al. 2024). A check to avoid detector small scale sensitivity issues⁴ potentially affecting the UVOT photometry was also performed.

$\lambda$ (Å)	$f$-0	$f$-1	$f$-2	$f$-3	$f$-4	$f$-5	$f$-6	$f$-7
10,200	−14.52378	−14.49784	−14.47881	−14.45454	−14.42847	−14.39924	−14.36778	−14.33206
10,180	−14.53819	−14.51207	−14.49217	−14.46784	−14.44060	−14.41209	−14.37972	−14.34315
10,160	−14.53943	−14.51288	−14.49213	−14.46828	−14.44113	−14.41154	−14.37821	−14.34210
10,140	−14.55275	−14.52525	−14.50488	−14.48021	−14.45388	−14.42401	−14.38964	−14.35441
10,120	−14.55215	−14.52438	−14.50430	−14.47896	−14.45343	−14.42336	−14.38902	−14.35524
10,100	−14.55647	−14.52956	−14.50961	−14.48412	−14.45902	−14.42974	−14.39513	−14.36277
10,080	−14.55812	−14.53196	−14.51187	−14.48719	−14.46222	−14.43297	−14.39914	−14.36655
10,060	−14.55376	−14.52878	−14.50821	−14.48373	−14.45821	−14.43052	−14.39743	−14.36534
10,040	−14.55684	−14.53181	−14.51184	−14.48703	−14.46226	−14.43534	−14.40252	−14.37078
10,020	−14.55390	−14.52987	−14.50967	−14.48493	−14.46077	−14.43388	−14.40202	−14.37081
10,000	−14.55453	−14.53047	−14.51054	−14.48618	−14.46174	−14.43522	−14.40359	−14.37391

Notes. The full table is available only electronically via CDS, a small portion is shown here for guidance on its form and content.

Photometric results in the UVOT Vega-system are reported in Table 5 together with those of X–ray pointings: from the latter no emission was detected in the 0.3–10 keV range for any of the sources; 3-$\sigma$ flux limits were determined using the procedure of Gehrels [50] with the exception of GaSS 2-55 for which the X–ray flux limit was evaluated through a more conservative approach using the xspec package v. 12.10.0 (Arnaud [8]) due to the high background during the XRT observation.

To define the sensitivity to accretion processes of observations in the UVOT bands of our program stars (cf. Sect 6.2 for details), we searched the Swift archive for serendipitous observations of field, high Galactic latitude, and non-variable red giants with accurate spectral classifications from Houk & Swift [62], which are sufficiently faint to avoid overloading the Swift detectors. We found 16 suitable red giants in the archive, which we supplemented with ToO observations of four additional objects (all of M spectral type). The results are summarized below in Table H.1.

				Swift UVOT		GALEX
star	SpType	date	UVW1	UVM2	UVW2	$B$ / NUV
HD 2059	K0 III	11/05/2017	$13.21 \pm0.04$	$17.31 \pm0.06$	$14.32 \pm0.04$	10.29 / $18.008 \pm0.031$
HD 88283	K1 III	16/05/2010	$13.02 \pm0.04$	$17.14 \pm0.09$	$14.22 \pm0.04$	10.23 / $18.229 \pm0.051$
HD 3704	K1/2 III	28/05/2017	$12.85 \pm0.04$	$17.11 \pm0.11$	$13.83 \pm0.04$	10.10 /
HD 89603	K2 III	23/11/2007	$12.21 \pm0.04$	$16.30 \pm0.05$	$13.42 \pm0.04$	9.48 / $17.318 \pm0.025$
HD 72211	K3 III	26/03/2005	$12.95 \pm0.04$	$17.67 \pm0.06$	$14.13 \pm0.04$	10.05 / $18.844 \pm0.074$
HD 91605	K3/4 III	08/07/2010	$12.83 \pm0.04$	$17.46 \pm0.10$	$14.09 \pm0.04$	10.01 / $18.601 \pm0.045$
HD 29943	K4 III	12/12/2021	$12.93 \pm0.04$	$17.29 \pm0.11$	$13.98 \pm0.04$	10.22 / $18.656 \pm0.046$
HD 1208	M0 III	20/04/20165	$12.98 \pm0.04$	$17.44 \pm0.09$	$13.95 \pm0.04$	10.07 / $18.801 \pm0.071$
HD 184438	M0 III	01/07/2019	$12.21 \pm0.04$	$16.57 \pm0.05$	$13.08 \pm0.04$	9.35 /
HD 214003	M1 III	19/09/20165	$12.18 \pm0.04$	$16.62 \pm0.07$	$13.10 \pm0.04$	9.26 / $18.129 \pm0.036$
HD 1263086	M2 III	20/03/2023	$13.36 \pm0.04$	$17.72 \pm0.15$	$14.27 \pm0.04$	10.37 / $19.045 \pm0.040$
HD 16756	M2 III	11/09/2013	$12.54 \pm0.04$	$16.97 \pm0.04$	$13.53 \pm0.04$	9.56 / $18.504 \pm0.067$
HD 7235	M2/3 III	26/01/2013	$13.00 \pm0.05$	$17.47 \pm0.04$	$14.10 \pm0.04$	10.16 / $18.819 \pm0.076$
HD 139867	M2/3 III	24/01/2018	$11.93 \pm0.04$	$16.74 \pm0.08$	$12.73 \pm0.04$	9.00 /
HD 31518	M3 III	14/11/20095	$11.82 \pm0.04$	$16.13 \pm0.05$	$12.86 \pm0.04$	8.84 /
HD 1547646	M3 III	24/03/2023	$13.77 \pm0.04$	$18.50 \pm0.20$	$14.78 \pm0.04$	10.77 / $19.651 \pm0.119$
HD 66424	M5 III	06/05/2019	$12.93 \pm0.04$	$17.45 \pm0.08$	$13.82 \pm0.04$	10.00 / $18.657 \pm0.057$
HD 834556	M6 III	24/03/2023	$12.67 \pm0.04$	$16.80 \pm0.09$	$13.63 \pm0.04$	9.78 / $18.552 \pm0.063$
HD 953846	M7/8 III	20/03/2023	$13.20 \pm0.04$	$18.33 \pm0.17$	$14.06 \pm0.04$	10.33 / $19.941 \pm0.157$

Notes. $^{a}$ = stars that we observed as Target-of-Opportunity sources with Swift UVOT. $^{b}$ = stars with observations in W1 and/or W2 bands carried out on a different date, the quoted one being for the M2 band. $B$ band magnitude as listed by SIMBAD. GALEX NUV magnitudes are in the native AB system.