© The Authors 2023

1. Introduction

Brown dwarfs (BDs) are substellar objects that have an insufficient mass to maintain stable hydrogen thermonuclear burning (Kumar 1963; Burrows & Liebert 1993; Baraffe et al. 1995). Since the discovery of the first BD, Teide 1 in the Pleiades (Rebolo et al. 1995), and the first BD companion, Gliese 229 B (Nakajima et al. 1995), thousands of BDs have been identified. Characterizing BDs of different ages, temperatures, masses, and metallicities is critical for tracing substellar evolution paths and building ultracool atmospheric models. BDs are extremely faint because of their low surface temperatures and small radii. Large-scale deep optical and near-infrared (NIR) sky surveys, such as the Two Micron All Sky Survey (2MASS, Cutri et al. 2003; Skrutskie et al. 2006); the Sloan Digital Sky Survey (SDSS, York et al. 2000); the United Kingdom Infrared Telescope Infrared Deep Sky Survey (UKIDSS, Lawrence et al. 2007); the Wide-field Infrared Survey Explorer (WISE, Wright et al. 2010); and the Dark Energy Survey (DES, Abbott et al. 2018; Morganson et al. 2018; Flaugher et al. 2015), have been used to search for brown dwarfs.

Brown dwarfs are classified into different spectral types based on their spectral features. After the spectral type M, new type sequences L, T, and Y have successively been established over the last three decades. The effective temperature drops from 2400 K for the earliest L type to less than 300 K for the latest Y type. Kirkpatrick et al. (1999) and Martín et al. (1999) defined the L-type spectral class by noting the strengthening of alkali metal lines, the fading of oxide bands, and the enhancement of hydride and water bands with respect to M dwarfs. Burgasser et al. (2002) and Geballe et al. (2002) independently made robust T-type classifications based on strong methane bands in NIR spectra. Both the L and T type have been divided into sub-classes from 0.0 (‘early type’) to 9.5 (‘late type’). It is important to note that late-M and early-L dwarfs are mixtures of young massive BDs and old very low-mass dwarf stars (Dupuy & Liu 2017; Zhang 2019). Delorme et al. (2008), Cushing et al. (2011), and Kirkpatrick et al. (2011) found the first examples of Y dwarfs and proposed that ammonia absorption in the H band could be a characteristic of this class, which is cooler than T. So far, only tens of Y dwarfs have been discovered and this class extends only to the Y2 subclass.

Metal-poor dwarfs are also called subdwarfs. Their chemical composition remains pristine, except for the fusion of some light elements such as deuterium (Chabrier et al. 2000) and lithium for the most massive ones close to the substellar limit (Rebolo et al. 1992; Basri et al. 1996). They have been continuously fading and cooling down for billions of years with increasingly degenerate cores and no significant thermonuclear reactions. Thousands of M subdwarfs (Zhang et al. 2019) and dozens of L subdwarfs (Zhang et al. 2018) have been identified, but only a select number of T subdwarf candidates have been announced. The optical properties of T subdwarf candidates have not been explored yet and may be crucial for breaking the metallicity and temperature degeneracy, as has been done for M and L subdwarfs (Zhang et al. 2017; Lodieu et al. 2019).

In this paper, we present observations of ten metal-poor T subdwarf candidates that allowed us to obtain photometry in the z band for all of them and optical spectroscopy for two of them. Section 2 describes the observation protocols and data reduction processes. Sections 3–5 show the results of astrometry, photometry, and spectroscopy, as well as our analysis. Section 6 describes a new potential metallicity indicator, the z − W1 colour, and discusses possible changes required for the current T subdwarf classification. Section 7 provides a summary and discussion of the impact of this research and future applications.

2. Observations

Our sample consists of ten T subdwarf candidates from the literature that are observable from the Roque de los Muchachos Observatory on the island of La Palma (Spain). Burningham et al. (2010, 2014) discovered ULAS0926, ULAS1316, and ULAS1319 from the Large Area Survey (LAS) of UKIDSS based on their blue NIR colours. Their ultracool subdwarf status was confirmed by NIR spectroscopy. Pinfield et al. (2014) discovered the two late-T subdwarf candidates WISE0013 and WISE0833 from W2-only detection in WISE and classified them based on their halo kinematics, metal-poor NIR colours, and NIR spectroscopy. Greco et al. (2019) identified WISE0004 and WISE0301 from the NEOWISE proper motion survey and classified them as T subdwarfs after NIR spectroscopic follow-up. Meisner et al. (2020, 2021) identified WISE0422, WISE1553, and WISE2217 from the Backyard Worlds: Planet 9 Citizen Science Project, which visually selects the candidates with high proper motions. Target information is listed in Table 1.

2.1. Observation details

We collected z-band photometry of these candidates using the imaging mode of Optical System for Imaging and low-Intermediate-Resolution Integrated Spectroscopy (OSIRIS Cepa et al. 2000) on the 10.4 m Gran Telescopio Canarias (GTC) on La Palma. We requested a four-point dithering pattern with ten 50 s exposures at each point. Our weather constraints were seeing less than 0.9″, grey nights, and clear sky. In practice, the support astronomer adapted the numbers of exposures for some of the objects to the observing conditions, resulting in slightly different total integration times (see Table 2).

We also carried out spectroscopic observations using the OSIRIS long slit mode of the two brightest targets, WISE0004 and WISE0301, under the same weather constraints as for the imaging mode. Before the spectroscopic exposures, we took three 30 s acquisition images with the z-band filter and made use of them for z-band photometry. We used the R500R grism and 1.0″ slit, resulting in a resolving power R ≈ 350. We selected the parallactic angle to prevent the flux from being lost due to atmospheric refraction. We created two observing blocks for two nights, each with two spectra of 1800 s shifted along the slit by 10″, and for each observing block we observed a spectroscopic standard star using the same grism and slit width with and without the z-band filter. Zapatero Osorio et al. (2018) first introduced this method of using the z-filtered standard spectrum to eliminate the second-order contamination (approximately from 9600 Å to 9800 Å) from the blue light (from 4800 Å to 4900 Å), which is a problem for all the OSIRIS red grisms. Our targets have a negligible flux in the blue part so there is no second-order contamination.

We note that ULAS0926 and WISE0833 were observed after the CCD replacement of OSIRIS on 2022 December 12. The main differences are that the new camera, called OSIRIS+, has one monolithic CCD instead of two, and that it improves the sensitivity specially in the blue wavelength range (< 5500 Å). A summary of all observations is provided in Table 2.

2.2. Data reduction

We reduced the OSIRIS z-band imaging data using IRAF (Tody 1986). First, the sky frame of each position in the four-point dithering pattern was created by median-scaling and median-combining all the images at the other three positions, but two-thirds of the highest pixels were rejected, as we regarded the lowest to represent the optimal sky value. The corresponding sky frame was then subtracted from all the images with respect to each position. Finally all the sky-subtracted images were aligned and average-combined, but the two highest and two lowest pixels were rejected.

We reduced the OSIRIS spectroscopic data with PypeIt (Prochaska et al. 2020a, 2020b). PypeIt can generate calibration frames, subtract the sky, extract the object spectra, do the wavelength calibration, generate the sensitivity function and use it to do the flux calibration, coadd the spectra, and finally perform the telluric correction. We tuned PypeIt to find the object in the y-pixel region between 1200 and 1600, because our objects are red and have very little flux below 8000 Å.

The sensitivity function for instrumental response correction was generated from the standard star spectrum. We spliced the part before 9000 Å of the non-z-filtered standard spectrum and the part after 9000 Å of the z-filtered standard spectrum to obtain the standard spectrum covering the whole wavelength range without second-order contamination; the method is explained in Sect. 2.1.

We fitted a low-order polynomial to the observed spectrum. The 9150 Å to 9900 Å water absorption wavelength region of the spectra was masked in the fit. Then the pipeline used the Maunakea telluric grids to correct the spectra for the telluric contribution.

3. Astrometry

3.1. Object recognition

We created the world coordinate system (WCS) for the OSIRIS images using Astrometry.net (Lang et al. 2010), which extracts stars and solves the WCS by matching subsets of four stars to the pre-computed 4200 series index based on the 2MASS catalogue. We requested Astrometry.net to match only the field around the telescope’s pointing position within one degree.

A 2000-s exposure of the GTC ORISIS z band goes deep down to 24.5 mag (3-σ, 1.0″ seeing and dark condition), such that there are no counterpart z-band images with a similar depth, requiring us to exclude the contamination and identify the objects. Therefore, to avoid false recognition, we projected the expected position region of the objects in the OSIRIS images based on their proper motions and their associated errors, the positions and the epoch of the corresponding position, and the time differences between the previous epoch and the epoch our OSIRIS observation. Table 1 lists the proper motions from the literature. We drew a circle with a radius equal to the error centred on the projected position for each target.

We detected, beyond a doubt, WISE0004, WISE0013, WISE0301, WISE0422, WISE0833, and ULAS1316 at their projected positions. However, there are no detections at the projected positions of ULAS0926, ULAS1319, and WISE2217. For ULAS0926 and ULAS1319, the proper motions given by Burningham et al. (2010) are over-estimated because the epoch differences are too short; and for WISE2217, the projection was slightly displaced from the object. The objects are indicated by the red arrows in the bottom three cutouts in Fig. A.1.

3.2. Proper motion revision

For the latter three objects, we used the pixel positions with the centroid errors given by the IRAF task imcentroid and the WCS from Astrometry.net to get their coordinates in our observations. To be sure about the astrometric accuracy, we calculated the rms of the fit of the coordinate transformation using the index pixel positions and field pixel positions in X and Y in corr.fits files generated by Astrometry.net. The rms is the standard deviation of the mean of the pixel deviation of the matched reference stars (match_weight > 0.99). For the three objects, the rmss are at the sub-pixel level. We derived proper motions using two epochs: the first epoch in Table 1 and the second epoch of our observations. Murray et al. (2011) gave the first epoch position errors of ULAS0926 and ULAS1319; and CatWISE provided those for WISE2217. The results are given in Table 3.

4. Photometry

We performed aperture photometry on the reduced images of those objects using the Astropy package photutils. The aperture radius is fixed to be 1″, and the background residual after the sky subtraction is from the median value in an annulus of 3.5″ inner radius and 5.5″ outer radius. We did the same photometric measurements on an extra set of three to six nearby stars of Pan-STARRS magnitudes between 18 mag and 20.3 mag (AB system, Oke 1974; Chambers et al. 2016). We used their z-band magnitudes to find the zero points. We assumed there to be Poisson errors and the uncertainties from the zero-point fit. The Poissonian error of aperture count A from the average combination of N exposures with the same exposure time is . The final error is simply the square root of the quadratic sum of the Poissonian error and the zero-point uncertainty.

4.1. Two extra objects: WISE0414 and WISE1810

For reference sources, we added two extra extreme T subdwarf candidates WISEA J041451.67−585456.7 and WISEA J181006.18−101000.5, which were reported and spectroscopically characterized in the NIR by Schneider et al. (2020). They have metallicities [Fe/H] ≈ −1 dex and ≤ − 1 dex, respectively. Lodieu et al. (2022) collected GTC OSIRIS z-band photometry and spectroscopy allowing them to place improved constraints on the metallicity of WISE1810 (−1.5 ± 0.5 dex). We also found that WISE0414 has z-band detection in the DES (see Fig. A.2). The DES also recorded its magnitudes in the r band at 26.54 ± 1.39 mag, and in the i band at 25.71 ± 1.13 mag.

4.2. Correction for Pan-STARRS z_AB magnitude

The z filter of Pan-STARRS¹ and the z′ filter of SDSS used by OSIRIS² are very different; the latter in particular allows the flux beyond 9300 Å to pass through and the former does not. Moreover, the profiles the spectral energy distributions (SEDs) of T dwarfs are quite inclined around this wavelength. Thus we are obliged to apply an offset to the GTC OSIRIS magnitudes to get the magnitudes under the Pan-STARRS AB system.

We synthesized photometry and computed the offsets between the OSIRIS magnitude and the Pan-STARRS magnitude z_AB against the NIR spectral types of dwarfs based on two filter profiles and L dwarf optical templates (Kirkpatrick et al. 1999), the T dwarf optical templates (Burgasser et al. 2002, 2003a), and the Y dwarf theoretical models (Morley et al. 2014). The Y dwarf models have gravity log g = 4.5, sedimentation efficiency F_sed = 5, cloud cover h = 50%, and effective temperature T_eff from 450 K to 200 K. We converted the T_eff of the Y dwarf to spectral types (Schneider et al. 2015; Cushing et al. 2021). Not all the dwarf templates are metal-poor, but we consider that, within a 1000 Å bandwidth in the z band, the difference in the SED slopes are so small that we can also apply this to our metal-poor samples. Because WISE0414 was detected in DES, we repeated the procedure for the DES filters³. The DES z filter has higher transmissivity than the SDSS z′ filter in the red part. Table 4 lists the results.

We did a consistency check on the calculated offset values. The two brightest objects have detections in Pan-STARRS DR1 (PS1): WISE0004, which has been recently classified as a T2 or T4, and WISE0301, which has been classified as a T1 (Best et al. 2020). Both are early T-type dwarfs. The differences between Pan-STARRS AB magnitudes and the GTC magnitudes z_AB −  are 0.58 ± 0.04 mag and 0.47 ± 0.05 mag, respectively. We recalculated the offsets using their ORISIS spectra (presented in Sect. 5) and found 0.61 mag and 0.52 mag, respectively. If we assume an uncertainty of ±1.0 in the sub-class in the spectral type of these two objects, both results are consistent with Table 4, which are from 0.68 to 0.87 mag and from 0.43 to 0.72 mag, respectively. We computed the offset of WISE1810 using its OSIRIS spectra (obtained by Lodieu et al. 2022 and presented in Sect. 5). WISE1810 has z_AB −  = 0.54 mag, which is consistent in Table 4 with its spectral type estimation (0.41 to 0.68 mag; T0.0 ± 1.0, Schneider et al. 2020). We also computed the offset for the Y dwarf WISE J173835.53+273259.0 with the OSIRIS spectrum obtained by Martín (in prep.). It has an estimated effective temperature of 300−450 K (Kirkpatrick et al. 2011), which corresponds to Y0–Y2, although Kirkpatrick et al. (2011) classified it as a Y0 in the NIR. The offset turns out to be 0.17 mag, which is in line with the predicted results in Table 4 (−0.12 to 0.59 mag). The mid-T types reach the maximum offset around 0.9 mag for both the GTC and the DES.

For the two brightest objects, WISE0004 and WISE0301, we simply adopted their Pan-STARRS AB magnitudes. For WISE1810, we used the correction calculated using its spectrum. For the rest, we applied the offsets in Table 4 to all the objects according to their estimated NIR spectral types in Table 1. The uncertainty of the correction comes from both the uncertainty of the spectral type and the coarseness of the spectral-type step in Table 4. The final z_AB error is the square root of the quadratic sum of this uncertainty and the photometric error of . The photometry results are shown in Table 5.

4.3. z-band magnitudes and colour–colour diagrams

Table 5 shows the GTC z-band magnitudes and the offset-corrected Pan-STARRS AB magnitude z_AB as well as the Y, J, H, K, W1, and W2 magnitudes or magnitude limits for all the targets in our sample. The z_AB of WISE0414 was corrected from its DES z magnitude, and its estimated spectral type T0.0 ± 1.0, which is the same as WISE1810 (Schneider et al. 2020). We noticed that our GTC magnitude of WISE1553 is in agreement with 22.17 ± 0.21 mag (Nidever et al. 2018) from Mosaic3 with a z filter similar to that of the DES, of the Kitt Peak Nicholas Mayall Telescope (Dey et al. 2016). We took the offset between our and DES z into account (shown in Table 4). Our magnitude for ULAS0926 is consistent at the 1-σ level with the magnitude 22.16 ± 0.09 mag converted by Burningham et al. (2010) from the z magnitude of the ESO Multi-Mode Instrument (EMMI). For ULAS1316, we adopted the Y, J, H, and K photometric measurement at epoch 2006.41 rather than those at epoch 2010.24, because ULAS1316 gradually moved onto a background galaxy during that time and thus the photometry had more contamination. For the same reason, the W1 and W2 magnitudes are questionable. For ULAS1319, there is no W1 nor W2 magnitude because the WISE satellite was not able to resolve it from a nearby bright star WISE J131943.68+120907.0 (W1 = 12.74 mag).

Figure 1 shows the colour–colour diagrams of W1 − W2 versus z_AB − W1 and J − W2 versus z − W1 for all the objects, except ULAS1316 and ULAS1319, with extra 8508 M dwarfs, 800 L dwarfs, and 42 T dwarfs from the Pan-STARRS1 3π survey (Best et al. 2018); one Y0 dwarf WISEP J173835.52+273258.9 was discovered by Cushing et al. (2011) with both a W1 and W2 detection and was detected in z using GTC OSIRIS (Lodieu et al. 2013); there are 39 L subdwarfs with photometric errors less than 0.2 mag in z, J, W1, and W2 bands (Zhang et al. 2018).

Fig. 1. W1 − W2 vs zAB − W1 and J − W2 vs zAB − W1 colour–colour diagrams of all the T subdwarf candidates in Table 5, except ULAS1319, which has neither W1 nor W2 magnitudes, and ULAS1316 whose photometry was contaminated by a background galaxy. For WISE0833 and WISE0013, we used arrows to indicate the lower limit of the W1 magnitude. We also included normal M, L, and T sequences; sub-L dwarfs; and a Y dwarf. All the T dwarfs, the Y dwarf, and our objects have error bars. In the first diagram, the candidates clearly separate into three groups.

Kirkpatrick et al. (2011) showed observationally that the W1 − W2 colour is a very powerful aid to distinguishing solar-metallicity T and Y dwarfs from M and L dwarfs; the colour becomes redder monotonically towards later types, and the slope of the colour versus spectral type relation curve increases dramatically after passing the L-T transition point. The J − W2 colour, however, has some degeneracy among late-L and early-T dwarfs.

In the first W1 − W2 versus z_AB − W1 colour–colour diagram, all of our objects have W1 − W2 ≥ 0.79 and are split into three quite distinctive groups in z_AB − W1 colours: Group I is mixed with the normal T dwarf sequence, Group III has an extremely red z_AB − W1 colour > 6.5 mag, and Group II lies in between Group I and Group III. The separation between the three groups in the second J − W2 versus z_AB − W1 colour–colour diagram is not as clear as that in the first diagram. We explain in Sect. 6 that z − W1 can be a good metallicity indicator for objects with W1 − W2 colours similar to those of T dwarfs.

5. Spectroscopy

Our two brightest targets, WISE0004 and WISE0301, lie in Group I at the beginning of the T dwarf sequence in the W1 − W2 versus z_AB − W1 colour–colour diagram. In this section we present and discuss their optical spectra (shown in Figs. 2–4). To compare the T subdwarf with a normal T dwarf, we used three spectra with T standards from the Low Resolution Imaging Spectrograph (LRIS) on Keck I, the T1 standard SDSS J083717.31−000018.0, the T5 standard 2MASS J07554795+2212169, and the T8 standard 2MASS J04151954−0935066 (Burgasser et al. 2003a). We also included the GTC OSIRIS spectrum of WISE1810, which is from Group III (Lodieu et al. 2022) in the comparison, but we shall discuss it in detail in Sect. 6.

Fig. 2. Full optical spectra from 5800 Å to 10470 Å normalized at 9250 Å in a logarithmic and a linear scale of the two T subdwarf candidates and WISE1810, with the alkali atomic lines (vertical lines or pressure broadened V-shaped lines), molecular bands (horizontal lines), and band heads (vertical lines with a dash). In the logarithmic scale plot, we smoothed the parts below 8000 Å of these three spectra to not let the noise block our sight. We plotted optical spectra of the T1 standard SDSS0837, the T5 standard 2MASS0755, and the T8 standard 2MASS0415 (Burgasser et al. 2003a) for comparison.

Fig. 3. Optical spectra normalized at 9250 Å on a linear scale in the alkali metal line region (Cs, Na and K, two top panels), and the hydride band regions (FeH and CrH, two bottom panels) of the two T subdwarf candidates and WISE1810 compared with three T dwarf templates. We did not plot the spectrum of WISE1810 because of its low signal-to-noise ratio.

Fig. 4. GTC OSIRIS spectra of WISE0004, WISE0301, and WISE1810 together with three T dwarf templates (Keck LRIS spectra) in the water band spectral region and a laboratory spectrum of pure H2O gas.

5.1. Spectral slope

Figure 2 shows that in the optical, the spectral slopes of our two targets are quite similar to each other and resemble the normal early-T dwarf slope. We used the least-squares method to compare the two spectra with the entire T dwarf standard grid provided by Burgasser et al. (2003a). We used the wavelength range from 8000 Å to 9250 Å. We found that WISE0004 matches T2 best, and WISE0301 matches T1. This is in agreement with the spectral types estimated from the NIR spectra for WISE0004 (T2) and WISE0301 (T1) by Greco et al. (2019). Although Best et al. (2020) estimate a type T4 for WISE0004.

5.2. Hydrides

WISE0301 shows a CrH bandhead starting at 8611 Å that is just like the T1 dwarf standard, but WISE0004 does not show it. This absorption feature disappears from spectral classes T1 to T5. These effects are illustrated in the lower left panel of Fig. 3.

Both targets have deeper FeH absorption at 9896 Å than the T1 standard and are similar to the T5 standard. This can be seen in the lower right panel of Fig. 3. Neither object shows strong CrH at 9969 Å or FeH at 8692 Å just like their solar-metallicity counterparts.

5.3. Alkali atomic lines

The Na I and K I resonance doublets are known to be extremely broad in ultracool dwarfs (Martín et al. 1999). The red wing of the Na I feature in our two targets shapes the optical spectra from 5900 Å to 7500 Å as can be seen in the upper logarithmic-scale plot in Fig. 2 and the upper right linear-scale plot in Fig. 3. In the same figures, the strong K I resonance doublet at 7665 Å and 7699 Å forms V-shaped notches in both targets. These are as deep and wide as that of the T5 standard.

The extended absorption due to the Na I and K I resonance features render the signal-to-noise ratio (S/N) in the continuum very low in our spectra shortwards of 8000 Å and this prevent us from detecting or setting any significant limits to the presence of the Li I resonance doublet at 6708 Å (a trademark of substellar status) or the Rb I resonance doublet at 7800 Å and 7948 Å.

The Cs I resonance doublet at 8521.1 Å and 8943.5 Å is prominent. We used the IRAF programmes rspec and splot to convert the Pypeit fits table to fits image and then to measure the equivalent width (EW) of these two lines. The measurement of the second line is tricky because it is located within the CH₄ absorption band. The first line of WISE0004 has an EW of 8.9 Å and that of WISE0301 has an EW of 8.4 Å; the second line of WISE0004 has an EW of 8.4 Å and that of WISE0301 has an EW of 7.4 Å. Burgasser et al. (2003a) reported that for early- and mid-T dwarfs, both of the caesium lines have an EW of about 7 to 9 Å, and Lodieu et al. (2015) reported that the EW of the caesium 8521.1 Å line of normal T dwarf peaks in early-T dwarfs is about 8 Å. The two T subdwarfs have EWs comparable to the EWs of normal T dwarfs.

5.4. Water bands

We compared our GTC OSIRIS T subdwarf spectra and WISE1810 spectra with normal T dwarf standards and with a laboratory spectrum of pure H₂O gas (Martín et al. 2021). The normal T dwarfs show a H₂O bandhead that reaches the deepest level at the beginning of the band at 9280 Å and attenuates for longer wavelengths. The T subdwarfs have the deepest absorption feature at 9350 Å similar to the laboratory water spectrum. The two T subdwarfs also have a second absorption feature centred at 9450 Å that is also seen in the laboratory spectrum but not in the normal T dwarfs. Readers can refer to the comparison in Fig. 4.

To explain the similarity of the optical water bands in the T subdwarfs and the laboratory spectrum, as well as the difference from the normal T dwarfs, we speculate that the physical conditions in metal-poor atmospheres may favour the condensation of pure water molecules that interact more among themselves and less with other molecules and with atoms than in the case of metal-rich atmospheres.

6. Metallicity gradient

Figure 5 illustrates how the optical and NIR spectra change in Group I, II, and III. For the Group I object WISE0004, the optical and the NIR part are from this study and NASA IRTF (Greco et al. 2019), respectively. For the Group II object WISE1553, only the NIR spectrum is from KECK NIRES (Meisner et al. 2021). We smoothed its spectrum with a window size of 21 to match the spectral resolution of other spectra. The optical spectrum of the Group III object WISE1810 is from GTC OSIRIS (Lodieu et al. 2022), and the NIR part is from Palomar/TripleSpec (Schneider et al. 2020).

Fig. 5. Normalized optical and NIR spectra (from the z to the K band) of five representative metal-poor ultracool dwarfs: the sdT2 WISE0004 from Group I in the colour–colour diagram shown in Fig. 1 (the optical part from this research and the NIR part from Greco et al. 2019); the Z-class prototype WISE1810 from Group III (the optical part from Lodieu et al. 2022 and the NIR part from Schneider et al. 2020); the sdT4 WISE1553 from Group II in between Group I and Group III (the only NIR spectrum from Meisner et al. 2021); the sdL8 2MASS0645 (Zhang et al. 2018); and the esdL7 2MASS0532 (Burgasser et al. 2003b). The major absorption bands and atomic lines are marked.

As can be seen in Fig. 5, Group I keeps the three NIR flux peaks of normal T dwarfs at 1.25, 1.60, and 2.10 μm (Burgasser et al. 2002) at the same positions, as well as the peak at 1.08 μm in the optical far red. Group III has the optical far red peak at 1.00 μm because of very broad triangular-shaped H₂O absorption that dominates in the J-band window. The NIR flux peaks are located at 1.3 μm, owing to the lack of CH₄ absorption in the J band, and 1.52 μm, possibly because of the lack of NH₃. The flux peak at 2.0 μm is flattened by the CIA of H₂.

The CH₄ absorption in the NIR is the trademark of T dwarfs and is shown by both T subdwarfs in Group I and II. Methane is what distinguishes T dwarfs from L dwarfs; in fact, T dwarfs are also known as methane dwarfs. However, the CH₄ features are very weak or completely missing in the Group III objects. The disappearance of methane is a clear sign of very low metallicity.

6.1. The z − W1 colour as a new metallicity indicator

There is a gradient of the spectral morphology between the groups: From Group I to Group III, the J-band H₂O absorption is getting stronger; the peak around 1.3 μm moves to the red; the H-band CH₄ absorption is getting weaker; the weaker NH₃ absorption makes the third peak and the fourth peak move bluewards; and the H and K bands get more suppressed by the CIA of H₂.

The metallicity estimation of T subdwarfs strongly relies on atmospheric modelling and it is very hard to consider developing, at the present time, a quantitative relationship between the physical observables and the metallicity values with such a small sample. The spectral morphology, however, can show the metallicity trend. Constructing a qualitative metallicity indicator based on this trend is feasible.

Group III objects have an estimated and approximate metallicity between −1.5 and −1.0 dex (Schneider et al. 2020; Lodieu et al. 2022). Meisner et al. (2021) reported that WISE1553 has a metallicity [Fe/H] ≈ −0.5 according to the NIR spectroscopy and the PHOENIX atmospheric models (Hauschildt & Baron 1999; Allard et al. 2013) extending to low metallicity (Gerasimov et al. 2020), but [Fe/H] ≲ − 1.5 according to the broad-band photometry and the LOWZ model of the author. Group I was classified as a T subdwarf but the exact metallicity is unknown. According to the most similarity between the normal T dwarf spectra and Group I spectra, we think that they have a metallicity close to the normal T dwarfs, that is −0.5 < [Fe/H] < 0.

Overall, we observed a general decreasing trend of metallicity from 0 to −1.5 dex from Group I to Group III that coincides with the increase of the z − W1 colour from 5 mag to 7 mag. Therefore, we argue that the robustness of the z − W1 colour might be a qualitative metallicity indicator for ultracool dwarfs with a W1 − W2 colour redder than 0.8 mag, which have temperatures lower than that of normal L dwarfs.

Cushing et al. (2005) showed there is a strong methane absorption in the MIR from 3.0 μm to 3.8 μm (fully covered by the W1 band) when the T dwarf sequence starts. The reddening of the z − W1 colour from Group I to Group III can be attributed to the attenuation of the major methane absorption in the W1 band. We also expect that there should be a limit for this indicator when the metallicity becomes too low and the W1 band is no longer affected by methane. Further MIR observations are needed to support this hypothesis and determine the lowest metallicity at which this indicator can still be valid.

6.2. Subdwarfs poor in methane

Two objects in Group III, WISE1810 and WISE0414, were tentatively designated as esdT0.0 ± 1.0 (Schneider et al. 2020). However, the lack of methane absorption features at the NIR wavelengths raises the question on the spectral classification of these objects which might require a revision once JWST MIR becomes available.

Schneider et al. (2020) also noticed that the CIA of hydrogen shapes the spectra of WISE1810 and WISE0414 in a way that resembles the extreme L subdwarfs in the H and K bands, indicating a similar metallicity to the extreme L subdwarf. Indeed, Lodieu et al. (2022) assigned a metallicity [Fe/H] ≈ −1.5 to WISE1810, which falls in the metallicity range of the extreme L subdwarfs, [Fe/H] from −1.0 down to −1.7 (Zhang et al. 2017). To answer the question that if Group III objects are in fact late-L subdwarfs, we show another comparison in the lower part of Fig. 5. The sdL8 2MASS J06453153−6646120 represents an L subdwarf (both the optical and the NIR part were taken by Zhang et al. 2018 using X-shooter on the VLT, Chile), together with the esdL7 2MASS J05325346+8246465 from Burgasser et al. (2003b) using LRIS and NIRSPEC on Keck I and II.

First of all, Group III objects have a redder W1 − W2 colour, not mixing with any L dwarfs or L subdwarfs in the W1 − W2 versus z − W1 colour–colour diagram. Second, The hydride bands, especially FeH, become stronger when metallicity decreases (Burgasser et al. 2007; Zhang et al. 2018), as seen from the comparison between sdL8 and esdL7 in Fig. 5. However, Group III objects do not show any signs of hydride in their optical or NIR spectra. Thirdly, the L subdwarfs and extreme L subdwarfs show very little H₂O absorption, but Group III objects have very strong H₂O bands in both the z and the J bands. Group III objects also do not show the strong potassium K I lines in the J band, which appear strong in both late-L subdwarf and extreme subdwarf spectra.

7. Conclusions

The photometric and spectroscopic characterization of metal-poor BDs widens the parameter space over which substellar mass objects are investigated. We used GTC OSIRIS to obtain z-band photometry for ten T subdwarf candidates, and optical spectroscopy for two of them, namely WISE0004 and WISE0301. We downloaded z-band photometry of WISE0414 from the DES database.

We noticed that our targets segregated into three subgroups in the W1 − W2 versus z − W1 colour–colour diagram. Group I mixed with the T dwarf sequence from the literature; Group III is composed of only three objects (WISE1810, WISE0414, and WISE2217) and has extraordinarily red z − W1 colours. Group II is located between the other two groups.

We examined the optical spectra of two early-T subdwarfs in Group I (WISE0004 and WISE0301). Their optical spectra appear similar to those of early-T dwarfs that are dominated by alkali metals and hydrides. We inferred that these two early-T subdwarfs do not have a very low metallicity. Nevertheless, they have a deeper water absorption feature in the optical. This water absorption feature in T subdwarf spectra is more similar to the water transmission absorption spectra obtained in the laboratory compared with normal T dwarf spectra.

Based on the peculiar position of these T subdwarf candidates in the colour–colour diagram, we compared the spectra of the early-T subdwarf WISE0004 in Group I, mid-T subdwarf WISE1553 in Group II, WISE1810 in Group III, the late-L subdwarf 2MASS0645, and the late-L extreme subdwarf 2MASS0532. We conclude that the z − W1 colour is qualitatively a good metallicity indicator for objects with temperatures comparable to that of T dwarfs.

The spectroscopy of the late-type objects in Group II, WISE0422, is needed in the future to test the robustness of the z − W1 colour as a metallicity indicator for cooler objects, as well as the spectroscopy of the third object in Group III, WISE2217, to confirm its metal-poor nature. Although, it will be extremely challenging (J = 20.66 mag).

The spectral classification of Group III objects might require a revision because of the weak or absent methane absorption at NIR wavelengths, a key molecular feature to classify T-type dwarfs. We anticipate that new JWST MIR observations will bring new insight into the spectral classification of metal-poor BDs. An option could be using new letters to classify spectra that are different to those previously classified with the letters L, T, and Y. We note that the letters H and Z are still available.

More objects of these three groups and those filling the gaps between them need to be discovered to improve the statistical significance of the connection between the z − W1 colour and metallicity. We also need more accurate metallicity measurements of these objects to be able to establish this relationship. It is promising that this will be achieved in the near future. A quantum leap is expected in the numbers of substellar objects that can be identified with the advent of deep large area surveys such as the Euclid space mission and Legacy Survey of Space and Time (LSST) at the Vera Rubin Observatory. Thousands of T dwarf slitless low-resolution NIR spectra are expected from the Euclid surveys (Martín et al. 2021), and LSST will become a powerful search engine for Group III objects and those even more metal-poor, expanding more than 1000 times of the space volume in the solar vicinity that we have explored so far⁴.

Acknowledgments

Funding for this research was provided by the European Union (ERC, SUBSTELLAR, project number 101054354) and the Agencia Estatal de Investigación del Ministerio de Ciencia e Innovación (AEI-MCINN) under grant PID2019-109522GB-C53. Based on observations made with the Gran Telescopio Canarias (GTC), installed at the Spanish Observatorio del Roque de los Muchachos of the Instituto de Astrofísica de Canarias, on the island of La Palma. This work is (partly) based on data obtained with the instrument OSIRIS, built by a Consortium led by the Instituto de Astrofísica de Canarias in collaboration with the Instituto de Astronomía of the Universidad Autónoma de México. OSIRIS was funded by GRANTECAN and the National Plan of Astronomy and Astrophysics of the Spanish Government. This research has made use of data provided by Astrometry.net. This work has used the Pan-STARRS1 Surveys (PS1) and the PS1 public science archive that have been made possible through contributions by the Institute for Astronomy, the University of Hawaii, the Pan-STARRS Project Office, the Max-Planck Society and its participating institutes, the Max Planck Institute for Astronomy, Heidelberg and the Max Planck Institute for Extraterrestrial Physics, Garching, The Johns Hopkins University, Durham University, the University of Edinburgh, the Queen’s University Belfast, the Harvard-Smithsonian Center for Astrophysics, the Las Cumbres Observatory Global Telescope Network Incorporated, the National Central University of Taiwan, the Space Telescope Science Institute, the National Aeronautics and Space Administration under Grant No. NNX08AR22G issued through the Planetary Science Division of the NASA Science Mission Directorate, the National Science Foundation Grant No. AST-1238877, the University of Maryland, Eotvos Lorand University (ELTE), the Los Alamos National Laboratory, and the Gordon and Betty Moore Foundation. This publication makes use of data products from the Wide-field Infrared Survey Explorer, which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technology, funded by the National Aeronautics and Space Administration. This project used public archival data from the Dark Energy Survey (DES). Funding for the DES Projects has been provided by the US Department of Energy, the US National Science Foundation, the Ministry of Science and Education of Spain, the Science and Technology Facilities Council of the United Kingdom, the Higher Education Funding Council for England, the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign, the Kavli Institute of Cosmological Physics at the University of Chicago, the Center for Cosmology and Astro-Particle Physics at the Ohio State University, the Mitchell Institute for Fundamental Physics and Astronomy at Texas A&M University, Financiadora de Estudos e Projetos, Fundação Carlos Chagas Filho de Amparo ã Pesquisa do Estado do Rio de Janeiro, Conselho Nacional de Desenvolvimento Científico e Tecnológico and the Ministério da Ciência, Tecnologia e Inovação, the Deutsche Forschungsgemeinschaft and the Collaborating Institutions in the Dark Energy Survey. The Collaborating Institutions are Argonne National Laboratory, the University of California at Santa Cruz, the University of Cambridge, Centro de Investigaciones Enérgeticas, Medioambientales y Tecnológicas-Madrid, the University of Chicago, University College London, the DES-Brazil Consortium, the University of Edinburgh, the Eidgenössische Technische Hochschule (ETH) Zürich, Fermi National Accelerator Laboratory, the University of Illinois at Urbana-Champaign, the Institut de Ciències de l’Espai (IEEC/CSIC), the Institut de Física d’Altes Energies, Lawrence Berkeley National Laboratory, the Ludwig-Maximilians Universität München and the associated Excellence Cluster Universe, the University of Michigan, the National Optical Astronomy Observatory, the University of Nottingham, The Ohio State University, the OzDES Membership Consortium, the University of Pennsylvania, the University of Portsmouth, SLAC National Accelerator Laboratory, Stanford University, the University of Sussex, and Texas A&M University. Based in part on observations at Cerro Tololo Inter-American Observatory, National Optical Astronomy Observatory, which is operated by the Association of Universities for Research in Astronomy (AURA) under a cooperative agreement with the National Science Foundation. This work made use of Astropy (http://www.astropy.org): a community-developed core Python package and an ecosystem of tools and resources for astronomy (Astropy Collaboration 2013, 2018, 2022). We appreciate the referee report for providing useful and insightful comments.

References

Appendix A: z-band images of the targets

Fig. A.1. 45′×45′ GTC OSIRIS z-band images of the fields of the all targets with the conventional direction (up is north and left is east). The blue dots are the object positions at the previous epochs published by other authors (Table 1), and the red ellipses are the projected positions where those targets are supposed to be on the dates of the observations in Table 2 according to their proper motions from the literature in Table 1. The two semi-axes of the ellipses show the errors of the proper motions in RA and Dec. WISE0004, WISE0013, WISE0301, WISE0422, WISE0833, ULAS1316, and WISE1553 were unambiguously detected within the error ellipses at their projected positions. However, for ULAS0926 and ULAS1319, we found that the proper motions given in the literature are overestimated; and for WISE2217, the proper motion is deviated. The red arrows indicate the positions at which the three sources should be. We recalculated their proper motions according to these positions and we listed the results in Table 3.

Fig. A.2. 45′×45′ DES z-band field of WISE0414 with the conventional direction (up is north and left is east). The red dot is the position WISE0414 in the VHS J band, whose observation epoch is only 27 days before the DES observation epoch.

All Tables

All Figures

Fig. 1. W1 − W2 vs zAB − W1 and J − W2 vs zAB − W1 colour–colour diagrams of all the T subdwarf candidates in Table 5, except ULAS1319, which has neither W1 nor W2 magnitudes, and ULAS1316 whose photometry was contaminated by a background galaxy. For WISE0833 and WISE0013, we used arrows to indicate the lower limit of the W1 magnitude. We also included normal M, L, and T sequences; sub-L dwarfs; and a Y dwarf. All the T dwarfs, the Y dwarf, and our objects have error bars. In the first diagram, the candidates clearly separate into three groups.

In the text

Fig. 2. Full optical spectra from 5800 Å to 10470 Å normalized at 9250 Å in a logarithmic and a linear scale of the two T subdwarf candidates and WISE1810, with the alkali atomic lines (vertical lines or pressure broadened V-shaped lines), molecular bands (horizontal lines), and band heads (vertical lines with a dash). In the logarithmic scale plot, we smoothed the parts below 8000 Å of these three spectra to not let the noise block our sight. We plotted optical spectra of the T1 standard SDSS0837, the T5 standard 2MASS0755, and the T8 standard 2MASS0415 (Burgasser et al. 2003a) for comparison.

In the text

	Fig. 3. Optical spectra normalized at 9250 Å on a linear scale in the alkali metal line region (Cs, Na and K, two top panels), and the hydride band regions (FeH and CrH, two bottom panels) of the two T subdwarf candidates and WISE1810 compared with three T dwarf templates. We did not plot the spectrum of WISE1810 because of its low signal-to-noise ratio.
In the text

	Fig. 4. GTC OSIRIS spectra of WISE0004, WISE0301, and WISE1810 together with three T dwarf templates (Keck LRIS spectra) in the water band spectral region and a laboratory spectrum of pure H2O gas.
In the text

Fig. 5. Normalized optical and NIR spectra (from the z to the K band) of five representative metal-poor ultracool dwarfs: the sdT2 WISE0004 from Group I in the colour–colour diagram shown in Fig. 1 (the optical part from this research and the NIR part from Greco et al. 2019); the Z-class prototype WISE1810 from Group III (the optical part from Lodieu et al. 2022 and the NIR part from Schneider et al. 2020); the sdT4 WISE1553 from Group II in between Group I and Group III (the only NIR spectrum from Meisner et al. 2021); the sdL8 2MASS0645 (Zhang et al. 2018); and the esdL7 2MASS0532 (Burgasser et al. 2003b). The major absorption bands and atomic lines are marked.

In the text

Fig. A.1. 45′×45′ GTC OSIRIS z-band images of the fields of the all targets with the conventional direction (up is north and left is east). The blue dots are the object positions at the previous epochs published by other authors (Table 1), and the red ellipses are the projected positions where those targets are supposed to be on the dates of the observations in Table 2 according to their proper motions from the literature in Table 1. The two semi-axes of the ellipses show the errors of the proper motions in RA and Dec. WISE0004, WISE0013, WISE0301, WISE0422, WISE0833, ULAS1316, and WISE1553 were unambiguously detected within the error ellipses at their projected positions. However, for ULAS0926 and ULAS1319, we found that the proper motions given in the literature are overestimated; and for WISE2217, the proper motion is deviated. The red arrows indicate the positions at which the three sources should be. We recalculated their proper motions according to these positions and we listed the results in Table 3.

In the text

	Fig. A.2. 45′×45′ DES z-band field of WISE0414 with the conventional direction (up is north and left is east). The red dot is the position WISE0414 in the VHS J band, whose observation epoch is only 27 days before the DES observation epoch.
In the text