WorksPhysics and Astronomy
Ca II IRT activity indices from spectroscopy: a baseline method for photospheric contribution estimates
Abstract
The high-resolution spectroscopic observations using Radial Velocity Spectrometer offers a unique opportunity to study stellar Ca II infrared triplet (IRT) activity. A key step in estimating activity indices is subtracting the photospheric contribution. The canonical template-based method often fails because, for many stars, the observed line cores are deeper than the model predictions, which prevents valid activity measurements. In this work, we explore an alternative approach that constructs a baseline from less active stars as a function of effective temperature (and metallicity) to represent the photospheric contribution. Our results show that, for dwarf stars, compared with the canonical method, the baseline method not only yields far more objects with valid activity measurements, but also avoids many extremely low and unreliable values (e.g., log). The Ca II IRT activity indices derived from the baseline method correlate well with other activity proxies, confirming their reliability as activity indicators. Furthermore, they show a relation with stellar rotation, not only recovering the typical rotation–activity relation, but also revealing fine structures that may be associated with different regimes in gyrochronology. Our findings suggest that the baseline method can serve as a robust alternative for deriving accurate activity estimates, and will be particularly useful for obtaining activity indices for tens of millions of stars with the upcoming Data Release 4.
Introduction
Magnetic activity of late-type stars provides insight into magnetic energy release processes operating at different atmospheric layers, which are observable as different phenomena (e.g., star spots, chromospheric plages, ultraviolet emission, and X-ray emission) across the electromagnetic spectrum [e.g. 17, 56, 61, 55]. One of the best-known measures of magnetic activity is the Mount-Wilson S-index () and the derived , calculated using the flux in the center of the Ca II H and K lines. The Ca II infrared triplet (Ca II IRT) lines at 8498, 8542, 8662 Å share the upper level of the H and K transitions and closely resemble them in many respects, indicating that they can also serve as chromospheric activity indicators (Chmieleski 2000; Montes et al. 2000; Busà et al. 2007).
Gaia mission performed high-resolution spectroscopic observations with the Radial Velocity Spectrometer (RVS), covering a wavelength range of 8450–8720 Å at a resolution of , for 32.2 million stars brighter than mag [50]. Considering that activity indices calculated from spectra are strongly affected by spectral resolution and that low-resolution spectra require further calibration [24], these high-resolution spectra provide a great opportunity to perform accurate Ca II IRT activity measurements. Lanzafame et al. [33] presented stellar activity indices derived from the Ca II IRT lines for two million stars in the Milky Way, using both published and unpublished RVS spectra.
Generally, the calculation of Ca II IRT activity indices involves two steps: integration of the flux in the cores of the Ca II IRT lines and removal of the photospheric contribution. Previous studies have developed different methods to perform these two steps. For example, Martin et al. [38] used an integration range of 1 Å around the core of each of the triplet lines, while Fuhrmeister et al. [19] and Lanzafame et al. [33] adopted a 0.5 Å and 1.5 Å interval, respectively. Busà et al. [8] even used a variable integration range, from 0.1 Å in slow-rotating, non-active stars up to 6 Å in fast rotators and/or very active stars. For the photospheric contribution, previous studies have employed local thermodynamic equilibrium (LTE) [33] or non-LTE (NLTE) [1, 8, 38, 28] synthetic spectral models. The resulting quantity after subtracting the photospheric contribution can be called excess equivalent width (). In addition to these two steps, Lanzafame et al. [33] further converted the excess to to express the flux excess over the radiative equilibrium approximation in the Ca II IRT cores in bolometric units.
However, previous methods for estimating the photospheric contribution have some disadvantages. The most significant issue is that, for many stars, the observed line cores are deeper than those predicted by the parameter-matched templates. Yang et al. [63] interpreted this discrepancy as arising primarily from the omission of chromospheric structure in templates, with a secondary contribution from NLTE effects. As a result, a large proportion of stars are prevented from yielding meaningful activity measurements.
In this study, we introduced a simple method to estimate the photospheric contribution (as well as the chromospheric emission unrelated to magnetic activity) using a baseline constructed from less active stars, which was also mentioned by Dempsey et al. [15]. We adopted two approaches to build this baseline, with and without accounting for metallicity, since metallicity can influence the emission levels of metal lines. The paper is organized as follows. Section 2 describes the dataset and the methods for estimating the Ca II IRT activity indices. Section 3 presents the results and discussion, covering the comparison of indices from different methods, the correlation between Ca II IRT and other activity proxies, and the rotation–activity relation derived from Ca II IRT activity. Finally, a summary is given in Sect. 4.
Dataset and method
Gaia Data Release 3 (DR3) has published high-resolution spectra (i.e., rvs_mean_spectrum) for 999,645 stars. These mean spectra were generated from multiple individual observations, each of which was first normalized to the pseudo-continuum and shifted to the rest frame before combination. We downloaded all these spectra from the Gaia archive data website2 By visual inspection, we found that the continua of some spectra are not well normalized. Therefore, we re-normalized all spectra before the subsequent analysis. We adopted laspec.normalization.normalize_spectrum_spline for normalization, which uses a cubic smoothing spline to model the continuum.
Atmospheric parameters
Atmospheric parameters are needed during the calculation of the activity indices. Gaia DR3 release presents two sets of stellar parameters: the General Stellar Parametrizer from spectroscopy (GSP-Spec) using high-resolution RVS spectra, and the GSP from photometry (GSP-Phot) using low-resolution BP/RP spectra. Lanzafame et al. [33] used the parameters from GSP-Spec by default, since the activity index was derived from the same data; when they were not available, the ones from GSP-Phot were used.
Here, we did not simply follow this approach. Instead, we first assessed which set of parameters is more reliable, with the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) parameter catalog (Zhang et al. [64]) as a reference. This catalog contains atmospheric parameters for 6.4 million stars released in LAMOST low-resolution DR9, derived with the DD-PAYNE method. We cross-matched the LAMOST and Gaia DR3 catalogs and selected the common sources for which LAMOST, GSP-Spec, and GSP-Phot all provided valid values. There are 97,381 common stars between LAMOST and GSP-Spec, and 91,004 common stars between LAMOST and GSP-Phot. Figure 1 presents the – diagram (also called Kiel Diagram), constructed using the common stars. The LAMOST and GSP-Phot parameters define a tighter sequence than the GSP-Spec parameters, especially for main-sequence stars. Figure 2 compares the atmospheric parameters from LAMOST with those from GSP-Spec and GSP-Phot for stars with values (in all three datasets) larger than 3.5 (e.g., main-sequence stars and some subgiants). The temperature difference between LAMOST and GSP-Spec is smaller than that between LAMOST and GSP-Phot, but with substantially larger scatter. The surface gravities from GSP-Phot are in good agreement with LAMOST, showing much better consistency than those from GSP-Spec. GSP uses [M/H] as its metallicity, while LAMOST uses [Fe/H]. We neglected this difference in the comparison. The metallicity difference between LAMOST and GSP-Spec is smaller than that between LAMOST and GSP-Phot, again with larger scatter. Therefore, we used a hybrid combination of parameters from GSP-Spec and GSP-Phot: the temperature and values were taken from GSP-Phot, while the metallicity values were taken from GSP-Spec. Specifically, the parameters teff_gspphot, logg_gspphot, and mh_gspspec in Gaia DR3 catalog were used for stellar classification (i.e., dwarf or giant) and for baseline estimation. A total of 437,782 stars (44%) have valid measurements for all three parameters. Using these parameters, we classified the stars into dwarfs and giants, as shown in Figure 3.

Figure 1. Kiel diagram using different parameter sets: LAMOST (left panel), Gaia GSP-Spec (middle panel), and GSP-Phot (right panel).

Figure 2. Comparison of atmospheric parameters between LAMOST and GSP-Spec (top panels), and between LAMOST and GSP-Phot (bottom panels) for stars with values (in all three datasets) larger than 3.5.

Figure 3. Kiel diagram of our sample. Blue, green, and red lines mark the separation between dwarfs (blue points) and giants (plum points). The light gray points are excluded from the sample.
Sample cleaning
We removed binaries and other contaminants, such as galaxies and active galactic nuclei (AGNs), by cross-matching our sample with several existing catalogs. First, we collected catalogs with binary classifications from light curve database, including Catalina (Drake et al. [16]), All-Sky Automated Survey for Supernovae (e.g., Jayasinghe et al. [32]; Christy et al. [13]), General Catalog of Variable Stars (Samus et al. [48]), Zwicky Transient Facility (Chen et al. [11]), Wide-field Infrared Survey Explorer (Chen et al. [10]), Transiting Exoplanet Survey Satellite (TESS; Howard et al. [27]; Prša et al. 2022; Gao et al. [21]; Boyle et al. [7]), the Washington Double Star Catalog (Mason et al. [39]), the Ninth Catalog of Spectroscopic Binary Orbits (Pourbaix et al. [44]), and Gaia (Rimoldini et al. [47]). Second, we removed sources listed in the Gaia DR3 non-single source catalog (Gaia Collaboration et al. [20]). Third, stars with RUWE values (from Gaia DR3 catalog) larger than 1.4, which were possibly binaries, were removed. Finally, we cross-matched the sample with SIMBAD (Wenger et al. [59]) and excluded contaminants including binaries, galaxies, and AGNs. A matching radius of 3 arcsec was adopted for all cross-matching.
Calculation of and
is the normalized Ca II IRT luminosity index after subtracting the photospheric radiation contribution. The activity indices of each Ca II IRT line are denoted as , , and , respectively.
First, we calculated the of each Ca II IRT line as follows,
The integration range was set to around the line centers of the three spectral lines (i.e., , , and ) following Martin et al. [38] and Huang et al. [28]. Here, was taken as the median value of the continuum, defined over regions located 10–30 from the line center on both sides of each line. Using or simply integrating in previous studies are equivalent, since the constant unity cancels out when the photospheric contribution is subtracted.
Second, we used three methods to estimate the photospheric contribution: (1) the canonical method using template matching Andretta et al. [1]; (2) a baseline constructed from less active stars as a function of temperature Dempsey et al. [15]; (3) a baseline constructed as a function of both temperature and metallicity. Unlike the canonical method, the baseline approach is less suitable for giant stars, because giants exhibit a wide range of values, which can significantly affect the depth of the spectral lines. In this paper, we therefore applied the baseline methods only to dwarfs, while the template method was used for both dwarfs and giants.
In Method 1, we used the PHOENIX model Husser et al. [29] to calculate the photospheric contribution (). We built a model grid of for between 2500 and 8000 K in steps of 100 K, between 3 and 7 in steps of 0.5, [Fe/H] between −2.0 and 1.0 in steps of 0.5, and between 0 and 50 km s in steps of 5 km s. For each star, we adopted the parameter in the Gaia DR3 catalog as the observed projected rotational velocity. For stars with observational parameters within the grid, we applied a four-dimensional interpolation to calculate .
In Methods 2 and 3, we built the baselines using the observed values. The baseline constructions for these two methods are shown in Figure 4 and Figure 5, respectively. We selected the top 5% of sources with the highest values (i.e., less active) and adopted their median as the baseline. Then, a grid of basal values, corresponding to , was constructed as a function of temperature for Method 2 and of temperature plus metallicity for Method 3. Due to the lack of sufficient stars in some grid cells, the baseline is only defined over a narrow temperature range. For example, in the metallicity bin of 0.5 < [M/H] < 1.0, baseline estimates are only available for stars with temperatures between ∼4500 and ∼6000 K (Figure 5). For stars with observational parameters within the grids, we applied one-dimensional and two-dimensional interpolation to calculate for Methods 2 and 3, respectively.

Figure 4. Baseline construction (Method 2) for each Ca II IRT line, for dwarfs (top panels) and giants (bottom panels), using less active stars as a function of effective temperature (). , , and represent values of Ca II 8498, 8542, and 8662 Å lines, respectively. Thick red lines represent the baselines. For comparison, blue, green, and purple lines represent the values calculated from PHOENIX models with [Fe/H] values of −0.5, 0, and 0.5, respectively. For dwarfs, solid and dashed lines correspond to and 4.0, respectively; for giants, solid and dashed lines correspond to and 2.5, respectively.

Figure 5. Baseline construction (Method 3) for each Ca II IRT line for dwarfs, using less active stars as a function of effective temperature () and metallicity ([M/H]). Thick lines denote the baselines. For comparison, black and cyan lines represent the EW values calculated from PHOENIX models with [Fe/H] values corresponding to lower and upper boundaries of each metallicity bin, respectively. Solid and dashed lines correspond to log = 4.5 and 4.0, respectively.
The photospheric contribution derived from template matching using PHOENIX models is notably higher than the baselines (Figure 4 and Figure 5), indicating that many stars have observed line depths deeper than those predicted by the templates (i.e., the observed line fluxes are lower), supporting recent findings of Yang et al. [63].
Third, we calculated the excess related to chromospheric activity for each Ca ii IRT line following
The activity index was then calculated following
where was estimated as
The continuum flux was estimated using the PHOENIX synthetic spectra. For each source, we selected the synthetic spectrum that best matches its effective temperature, surface gravity, and metallicity, performed a continuum fit, and adopted the continuum flux at each Ca ii IRT line core as .
Finally, we calculated by averaging , , and . All and values derived from Methods 1, 2, and 3 are listed in Table 1, Table 2, and Table 3, respectively.
| Gaia DR3 | (K) | [M/H] | vbroad (km s) | |||||
| 111600430360448 | ... | |||||||
| 133732397478144 | ... | ... | ||||||
| 143421843569792 | ||||||||
| 311952064880256 | ||||||||
| 397718267299968 | ... | ... | ||||||
| 513884247145344 | ||||||||
| 1781247492276480 | ... | ... | ||||||
| 2662295608063872 | ... | ... | ||||||
| 2787674293549696 | ... | ... | ||||||
| 5659667384707456 |
Table 1. and measurements for both dwarfs and giants from Method 1. Notes. The parameters , , [M/H], and vbroad are taken from teff_gspphot, logg_gspphot, mh_gspspec, and vbroad in Gaia DR3 catalog. This table is available in its entirety in machine-readable and Virtual Observatory (VO) forms in the online journal. A portion is shown here for guidance regarding its form and content.
| Gaia DR3 | (K) | [M/H] | |||||
| 9281425163264 | |||||||
| 17149805245440 | |||||||
| 45664092292480 | |||||||
| 70476119170048 | |||||||
| 86105504493056 | |||||||
| 133732397478144 | |||||||
| 142730353190144 | |||||||
| 143421843569792 | |||||||
| 174959788042496 | |||||||
| 180697864082688 |
Table 2. and measurements for dwarfs from Method 2. Notes. The parameters , , and [M/H] are taken from teff_gspphot, logg_gspphot, and mh_gspspec in Gaia DR3 catalog. This table is available in its entirety in machine-readable and Virtual Observatory (VO) forms in the online journal. A portion is shown here for guidance regarding its form and content.
| Gaia DR3 | (K) | [M/H] | |||||
| 9281425163264 | |||||||
| 12545600306304 | ... | ||||||
| 17149805245440 | |||||||
| 45664092292480 | |||||||
| 70476119170048 | |||||||
| 86105504493056 | |||||||
| 133732397478144 | |||||||
| 142730353190144 | |||||||
| 143421843569792 | |||||||
| 174959788042496 |
Table 3. log and log measurements for dwarfs from Method 3.
Results and discussion
Comparison of among different methods
Figure 6 (top panels) compares the and values derived using Methods 1 and 3. As discussed above, the values derived from template matching are smaller than the observed values for a significant fraction of stars. As a result, not only do many stars fail to yield valid and measurements, but the derived values are also much smaller than those obtained with the baseline methods, leading to systematically lower and values.

Figure 6. Comparison of and values between Methods 1 and 3 (top panels) and between Methods 2 and 3 (bottom panels). The inset shows stars with temperatures lower than 4300 K.
At temperatures around 4000 K, the template-derived values closely match the baseline-derived values, with increasing difference toward both lower and higher temperatures (Figure 4 and Figure 5). Accordingly, for stars near 4000 K, the and values from Methods 1 and 3 are most consistent.
Figure 6 (bottom panels) presents a comparison of and values derived from Methods 2 and 3. It is clear that Method 2 yields higher values for metal-poor stars and lower values for metal-rich stars. This is expected, as the intrinsically shallower Ca ii IRT absorption lines in metal-poor stars would lead to an apparent enhancement in the measured activity (Soderblom et al. [52]). In Method 2, the baseline was constructed without considering metallicity; therefore, the metallicity effect remains imprinted in the activity values. In contrast, Method 3 incorporates both temperature and metallicity in the baseline construction, effectively weakening the metallicity effect. Such metallicity dependence has also been noted in many studies of Ca ii H&K activity (e.g., Lorenzo-Oliveira et al. [36]; Carvalho-Silva et al. [9]; Han et al. [24]).
With the same initial sample, Method 1 returns only 32,374 valid activity measurements (Table 1), including 2,526 giants and 29,848 dwarfs. By comparison, Methods 2 and 3 return 253,586 and 178,081 valid measurements for dwarfs (Table 2 and Table 3), respectively. Thus, our baseline method enables the construction of a much larger sample of activity measurements for statistical studies.
Figure 7 shows the Hertzsprung-Russell diagram for our sample dwarfs, which is dominated by F and G stars, with only a few M dwarfs. The Method 1 results (top panels) contain about 46% F, 46% G, 6% K, and 0.06% M stars. In comparison, the proportions for Method 2 (middle panels) are about 61%, 29%, 4%, and 0.4%, while those for Method 3 (bottom panels) are around 62%, 31%, 4%, and 0.3%. A group of possible unresolved binaries (or pre-main-sequence stars) lies above the main sequence. As activity increases (from left to right panels), our sample shifts to lower-mass stars.

Figure 7. Hertzsprung-Russell diagram of the sample stars, with colors representing the number of stars in each bin. Results for Methods 1, 2, and 3 are shown in the top, middle, and bottom panels, respectively. The left, middle, and right panels show stars with values in the 1%–25%, 25%–75%, and 75%–99% percentiles, respectively.
Comparison of with other proxies
The strength of magnetic activity can be traced by many proxies, including spots, flares, chromospheric emission (e.g., Ca ii H&K, H, and ultraviolet), and X-ray and radio emission. Generally, these proxies are correlated with each other (e.g., Güdel [22, 23]; Mamajek & Hillenbrand [37]; Stelzer et al. [55]; Wang et al. [58]).
We compiled a catalog of stars with () estimates derived from the literature, including ROSAT (Han et al. [26]), Chandra and XMM-Newton (Li et al. 2026), and eROSITA ([18]). Figure 8 shows the comparison of X-ray activity proxy with the and values derived from different methods. The Ca II IRT activity correlates positively with X-ray activity, with the results from Methods 2 and 3 showing tighter relations than those from Method 1. We carried out linear fits for each comparison, and the resulting best-fit parameters are given in Table 4. As can be seen, the uncertainties of fitted parameters for Methods 2 and 3 are much smaller than those for Method 1.

Figure 8. Comparison between from the literature and and derived from Method 1 (top panels), Method 2 (middle panels), and Method 3 (bottom panels). X-ray data were taken from ROSAT (green points), Chandra and XMM-Newton (purple), and eROSITA (orange) observations.
| Method | ||||||||
| m1 | ||||||||
| m2 | ||||||||
| m3 |
Table 4. Linear fit results of and as a function of : .
The number of common stars between the X-ray sample and Method 2 is larger than that with Methods 1 and 3, because the latter two require both teff_gspphot and mh_gspspec values, which are only available for 44% of the full sample. Moreover, Method 1 fails for many stars because the PHOENIX templates have shallower lines than the observed spectra. We also noticed that the stars uniquely covered by Method 2 (i.e., not by Methods 1 or 3) tend to show higher Ca II IRT and X-ray activities (e.g., and ). Further inspection of these active objects reveals that they all have effective temperatures lower than 6000 K.
We constructed a catalog of stars with measurements from the literature, based mainly on Boro Saikia et al. [6]. We noticed that some stars of Boro Saikia et al. [6] have different names or multiple values. For each source, we collected all available values and derived the median. Although Boro Saikia et al. [6] already included sources from Isaacson & Fischer [30], we found that some stars from the latter were missing. We therefore supplemented the measurements with sources from Isaacson & Fischer [30], as well as Isaacson et al. [31] and Sayeed et al. [51].
Figure 9 shows the comparison of the literature values with the and values derived from different methods. Overall, they are positively correlated. However, due to the small sample size and the scatter, we did not attempt to fit relations for these quantities.

Figure 9. Comparison between from the literature and and derived from Method 1 (top panels), Method 2 (middle panels), and Method 3 (bottom panels). Orange and green points are from Boro Saikia et al. [6] and other studies, respectively.
Comparison of with Gaia activity indices
Lanzafame et al. [33] presented a catalog of Ca II IRT activity indices (denoted as activityindex_espcs or ) for two million sources. They integrated the flux in the cores of each Ca II IRT line over a 1.5 Å window and removed the photospheric contribution using template matching based on stellar effective temperature, surface gravity, and metallicity. They also provided a conversion from into that depends on both temperature and metallicity (their Eq. 7), applicable for stars with metallicities between −1.0 to 1.0 and effective temperatures between 3000 to 7000 K.
We cross-matched our initial sample of one million sources with that catalog, and calculated values (from ) for about 0.18 million sources. Figure 10 (top panels) compares values with our values derived from Methods 1, 2, and 3. Since the method of Lanzafame et al. [33] is similar to our Method 1, the two agree well in the comparison. In contrast, the comparisons between and our Methods 2 and 3 show that is systematically lower, consistent with the behavior seen when comparing Method 1 with Methods 2 and 3 (Figure 6).

Figure 10. Top panels: Comparison of from the Gaia DR3 catalog with our values (derived from Methods 1, 2, and 3). Colors indicate the number of stars. Bottom panels: Comparison of from the Gaia DR3 catalog with the literature and values. The equation from Freund et al. [18] is ; the equation from Lanzafame et al. [33] is .
Different studies can yield different chromospheric activity measurements, since the data and procedures involved can differ. considerably, such as spectral resolution, normalization method, integration width, pseudocontinuum definition, and photospheric correction. However, both Lanzafame et al. [33] and our Method 1 include many stars with indices between −6.5 to −8.0.
About 33% of the Method 1 results (Table 1) fall in this range, compared with only 2% and 3% for Methods 2 and 3 (Table 2 and Table 3), respectively. We suggested that the discrepancy is driven primarily by different methods used to correct for the photospheric contribution. The very low values from Lanzafame et al. [33] were also noted by Freund et al. [18], who suggested that stars with may have unreliable indices.
We also cross-matched the values of the 0.18 million sources with the literature and data described in Section 3.2. The vs. diagram is similar to that shown by Freund et al. [18], and a simple linear function does not appear to fit the data well. For example, for , increases with , while for , remains nearly constant at a high level. Due to the small sample size, the vs. comparison shows notable scatter. In fact, Lanzafame et al. [33] also showed a large scatter between these two quantities. In their Fig. 7, many stars have values lower than , and exhibit poor correlation with .
Rotation–activity relation
Stellar rotation–activity relation is expected to shed light on magnetic dynamos operating in different stars. The canonical rotation–activity relation ([43, 61, 60]) is characterized by two distinct regimes. In the saturated part, rapid rotators maintain a constant activity level, which is thought to be driven by a convective dynamo. In the decay part, slow rotators show activity that declines with decreasing rotation, indicating an interface dynamo. In this section, we would investigate this relation using the Ca II IRT activity proxy.
We compiled stellar rotation periods from several catalogs ([40, 49, 46, 21]), spanning observations from the Kepler, K2, and TESS. When a star was listed in more than one catalog, we gave priority to the period determined from the higher-quality photometry (i.e., Kepler > K2 > TESS). To ensure reliable period measurements for the TESS sample ([21]), we selected only stars with both a high fitting threshold () and a high correct classification probability (). The Rossby number () is the ratio of the rotation period to the convective turnover time (). Here we used the Yale-Potsdam Stellar Isochrone grids ([53]) to calculate global , by performing a three-dimensional interpolation over effective temperature, surface gravity, and metallicity for each target.
Figure 11 shows the relation between the activity indices (, ) and the Rossby number. Only Method 2 reveals the saturation regime due to its large sample size. It is worth noting that different methods may yield different decay slopes in the decay regime. Figure 12 further shows the rotation–activity relation for different spectral types, based on the activity results from Method 2. Besides the canonical saturation and decay regimes, fine structures are also visible. The decay part does not follow a single slope; instead, it exhibits two distinct trends, especially for stars with temperatures between 4500 and 5700 K.
In gyrochronology, a main-sequence star evolves through a convective phase (younger, faster rotating, with a convective dynamo) and an interface phase (older, slower rotating, with an interface dynamo), separated by a transitional phase known as the gap [3, 4, 5]. Previous studies have linked the saturated and decay regimes in the rotation–activity relation to the convective and interface phases of gyrochronology, respectively [2, 61]. Recently, Yang et al. (2025) and Han et al. [25] identified an intermediate regime between the saturated and decay regimes in their rotation–activity relations and associated it with the gap phase, thereby establishing a one-to-one correspondence between the rotation–activity relation and gyrochronology. Furthermore, Han et al. (2026) reported a variable activity decay rate in the intermediate regime, implying a varying core–envelope coupling efficiency during early stellar evolution [34, 54, 14].
In Figure 12, the decay slope in the range differs from that at . We suggested the steeper-slope region of corresponds to the gap phase, while the regions before and after this range correspond to the convective and interface sequences, respectively. In addition, the shape of the relation closely resembles the findings of Mittag et al. [41], who found stellar rotation–activity cannot be described by a two-phase model and split their relation into four parts (their Fig. 7). Although we did not perform a quantitative fit of our relation, these results suggest the potential of using Ca ii IRT activity, derived with our baseline method, for detailed studies of stellar magnetic dynamos and associated physical processes.

Figure 12. Rotation–activity relation for stars in different bins, with (left panels) and (right panels) plotted against . Period measurements were from TESS (orange points), K2 (purple), and Kepler (green) observations. The two vertical dashed lines mark and .
Summary
In this work, we measured Ca ii IRT activity indices ( and ) using published Gaia RVS spectra with three different methods and compared the results. The procedure involves measuring the from the core flux of each Ca ii IRT line, subtracting the photospheric contribution, and converting the excess to activity index. The main difference among the three methods lies in how the photospheric contribution is treated. Method 1 follows the standard approach, using PHOENIX templates matched to effective temperature, surface gravity, and metallicity; Method 2 builds a baseline from less active stars as a function of effective temperature alone; Method 3 builds a baseline as a function of both effective temperature and metallicity.
Method 1 has the advantage of providing star-specific photospheric templates based on atmospheric parameters, but it often fails for stars whose observed line cores are deeper than the model predictions, preventing reliable activity estimates. Methods 2 and 3 are computationally efficient and straightforward to carry out. Method 2 requires only stellar temperature, enabling activity measurements for a large number of stars, although the resulting activity values are subject to metallicity effect. The limitation of Methods 2 and 3 is that they are less suitable for giant stars, where the large range of causes notable variations in line depth and width, making a baseline based solely on temperature and metallicity inadequate for these objects.

Figure 11. Rotation–activity relation constructed with Ca ii IRT activity indices (from Methods 1, 2, and 3). Period measurements were from TESS (orange points), K2 (purple), and Kepler (green) observations.
Comparing the activity results derived from the three methods, we found that Methods 2 and 3 not only return far more objects with valid and measurements, but also avoid many very low and unreliable measurements (e.g., ) seen in Method 1. In addition, the results from Method 2 are metallicity-dependent: they are lower than those from Method 3 for metal-rich stars, but higher for metal-poor ones.
We examined the relationship between the Ca II IRT activity indices and other activity indicators, including the Ca H&K emission () and X-ray emission (). In general, the Ca II IRT indices correlate positively with these proxies. Due to the larger sample size, Method 2 yields the widest range of activity levels and the tightest correlations.
The correlation between Ca II IRT activity and stellar rotation further confirms that and are reliable activity indicators. Clear saturation and decay regimes can be seen in the rotation–activity relation, particularly for the results from Method 2. Of particular interest is the presence of fine structures within the relation. The decay part appears to split into two regions: one in the range and the other at .
The former interval may correspond to the gap regime defined in gyrochronology.
A difference between our method and the canonical method is that our method not only estimates the photospheric contribution but also accounts for weak chromospheric emission unrelated to magnetic activity, such as acoustic-wave heating (Udnæs & Pereira 2025). However, one caveat is that for stars at low activity levels, the photospheric and non-magnetic chromospheric contributions complicate the physical interpretation and can lead to large uncertainty in activity calculation. Therefore, one should be cautious when applying any method to low-activity stars.
To date, fewer than 10,000 stars have high-resolution spectroscopic observations of Ca II H&K lines [e.g. 30, 6, 31]. Considering that the activity indices calculated from spectra are strongly affected by spectral resolution, the vast sample of high-resolution Ca II IRT spectra from Gaia offers an invaluable database for studying chromospheric activity. Our results sugNotes. The parameters , , and [M/H] are taken from teff_gspphot, logg_gspphot, and mh_gspspec in Gaia DR3 catalog. This table is available in its entirety in machine-readable and Virtual Observatory (VO) forms in the online journal. A portion is shown here for guidance regarding its form and content.
gest that until the template-based methods can provide a reliable estimate of the photospheric contribution to the Ca ii IRT lines, the baseline method can serve as a robust alternative for deriving accurate activity estimates. This approach will be especially useful for deriving Ca ii IRT activity indices from the upcoming Gaia DR4 data.
Acknowledgements. This work was supported by the National Natural Science Foundation of China (NSFC; grant Nos. 12588202/12673041/12273057), the National Key R&D Program of China (grant Nos. 2025YFF0511203/2025YFF0511200), the Strategic Priority Research Program of the Chinese Academy of Sciences (grant No. XDB1160302), the NAOC Project with No. E6517401, since Research Grants from the China Manned Space Project with No. CMS-CSST-2021-A08, and the New Cornerstone Science Foundation through the New Cornerstone Investigator Program and the XPLORER PRIZE. This work presents results from the European Space Agency (ESA) space mission Gaia. Gaia data are being processed by the Gaia Data Processing and Analysis Consortium (DPAC). This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France.
References
- [1]Andretta, V., Busà, I., Gomez, M. T., & Terranegra, L. 2005, A&A, 430, 669
- [2]Barnes, S. A. 2003a, ApJ, 586, L145
- [3]Barnes, S. A. 2003b, ApJ, 586, 464
- [4]Barnes, S. A. 2007, ApJ, 669, 1167
- [5]Barnes, S. A. & Kim, Y.-C. 2010, ApJ, 721, 675
- [6]Boro Saikia, S., Marvin, C. J., Jeffers, S. V., et al. 2018, A&A, 616, A108
- [7]Boyle, A. W., Bouma, L. G., & Mann, A. W. 2026, ApJS, 284, 75
- [8]Busà, I., Aznar Cuadrado, R., Terranegra, L., Andretta, V., & Gomez, M. T. 2007, A&A, 466, 1089
- [9]Carvalho-Silva, G., Meléndez, J., Ratham, A., et al. 2025, ApJ, 983, L31
- [10]Chen, X., Wang, S., Deng, L., de Grijs, R., & Yang, M. 2018, ApJS, 237, 28
- [11]Chen, X., Wang, S., Deng, L., et al. 2020, ApJS, 249, 18
- [13]Christy, C. T., Jayasinghe, T., Stanek, K. Z., et al. 2023, MNRAS, 519, 5271
- [14]Curtis, J. L., Agüeros, M. A., Matt, S. P., et al. 2020, ApJ, 904, 140
- [15]Dempsey, R. C., Bopp, B. W., Henry, G. W., & Hall, D. S. 1993, ApJS, 86, 293
- [16]Drake, A. J., Graham, M. J., Djorgovski, S. G., et al. 2014, ApJS, 213, 9
- [17]Frasca, A., Biazzo, K., Taş, G., Erdem, S., & Lanzafame, A. C. 2008, A&A, 479, 557
- [18]Freund, S., Czesla, S., Fuhrmeister, B., et al. 2025, A&A, 697, A230
- [19]Fuhrmeister, B., Czesla, S., Schmitt, J. H. M. M., et al. 2019, A&A, 623, A24
- [20]Gaia Collaboration, Arenou, F., Babusiaux, C., et al. 2023, A&A, 674, A34
- [21]Gao, X., Chen, X., Wang, S., & Liu, J. 2025, ApJS, 276, 57
- [22]Güdel, M. 2002, ARA&A, 40, 217
- [23]Güdel, M. 2004, A&A Rev., 12, 71
- [24]Han, H., Wang, S., Li, X., Zheng, C., & Liu, J. 2025, ApJ, 984, 2
- [25]Han, H., Wang, S., Yang, H., et al. 2026, ApJ, 1000, 208
- [26]Han, H., Wang, S., Zheng, C., et al. 2024, ApJS, 273, 8
- [27]Howard, E. L., Davenport, J. R. A., & Covey, K. R. 2022, RNAAS, 6, 96
- [28]Huang, X., He, Y., Bai, Z., et al. 2024, ApJS, 272, 6
- [29]Husser, T.-O., Wende-von Berg, S., Dreizler, S., et al. 2013, A&A, 553, A6
- [30]Isaacson, H. & Fischer, D. 2010, ApJ, 725, 875
- [31]Isaacson, H., Howard, A. W., Fulton, B., et al. 2024, ApJS, 274, 35
- [32]Jayasinghe, T., Kochanek, C. S., Stanek, K. Z., et al. 2018, MNRAS, 477, 3145
- [33]Lanzafame, A. C., Brugaletta, E., Frémat, Y., et al. 2023, A&A, 674, A30
- [34]Lanzafame, A. C. & Spada, F. 2015, A&A, 584, A30
- [36]Lorenzo-Oliveira, D., Freitas, F. C., Meléndez, J., et al. 2018, A&A, 619, A73
- [37]Mamajek, E. E. & Hillenbrand, L. A. 2008, ApJ, 687, 1264
- [38]Martin, J., Fuhrmeister, B., Mittag, M., et al. 2017, A&A, 605, A113
- [39]Mason, B. D., Wycoff, G. L., Hartkopf, W. I., Douglass, G. G., & Worley, C. E. 2001, AJ, 122, 3466
- [40]McQuillan, A., Mazeh, T., & Aigrain, S. 2014, ApJS, 211, 24
- [41]Mittag, M., Schmitt, J. H. M. M., & Schröder, K.-P. 2018, A&A, 618, A48
- [43]Pizzolato, N., Maggio, A., Micela, G., Sciortino, S., & Ventura, P. 2003, A&A, 397, 147
- [44]Pourbaix, D., Tokovinin, A. A., Batten, A. H., et al. 2004, A&A, 424, 727
- [46]Reinhold, T. & Hekker, S. 2020, A&A, 635, A43
- [47]Rimoldini, L., Holl, B., Gavras, P., et al. 2023, A&A, 674, A14
- [48]Samus, N. N., Kazarovets, E. V., Durlevich, O. V., Kireeva, N. N., & Pastukhova, E. N. 2017, AR, 61, 80
- [49]Santos, A. R. G., García, R. A., Mathur, S., et al. 2019, ApJS, 244, 21
- [50]Sartoretti, P., Marchal, O., Babusiaux, C., et al. 2023, A&A, 674, A6
- [51]Sayeed, M., Huber, D., Chontos, A., & Li, Y. 2025, AJ, 170, 212
- [52]Soderblom, D. R., Duncan, D. K., & Johnson, D. R. H. 1991, ApJ, 375, 722
- [53]Spada, F., Demarque, P., Kim, Y. C., Boyajian, T. S., & Brewer, J. M. 2017, ApJ, 838, 161
- [54]Spada, F. & Lanzafame, A. C. 2020, A&A, 636, A76
- [55]Stelzer, B., Marino, A., Micela, G., López-Santiago, J., & Liefke, C. 2013, MNRAS, 431, 2063
- [56]Strassmeier, K. G. 2009, A&A Rev., 17, 251
- [58]Wang, S., Bai, Y., He, L., & Liu, J. 2020, ApJ, 902, 114
- [59]Wenger, M., Ochsenbein, F., Egret, D., et al. 2000, A&AS, 143, 9
- [60]Wright, N. J. & Drake, J. J. 2016, Nature, 535, 526
- [61]Wright, N. J., Drake, J. J., Mamajek, E. E., & Henry, G. W. 2011, ApJ, 743, 48
- [63]Yang, X., Fu, X., Jian, M., et al. 2026, arXiv e-prints, arXiv:2604.14642DOI
- [64]Zhang, M., Xiang, M., Ting, Y.-S., et al. 2025, ApJS, 279, 5