2 Observations and data reduction

2.1 Near-IR observations

The near-IR spectra were obtained with the Australian National University 2.3 m Telescope at Siding Spring Observatory, using the cross-dispersed grism mode of the Cryogenic Array Spectrometer and Imager CASPIR. This camera operates at 70 K with a Santa Barbara Research Center CRC463 256$\times$256 InSb array maintained at 32 K (complete information is given in the CASPIR manual, McGregor 1994).

For each of the IJ, JH and HK grisms, five orders are imaged onto the CASPIR array. As can be seen from Table  $\ref{orders.tab}$, there is a significant overlap between orders and grisms. The spectra were acquired through the $1^{\prime \prime} \times 15^{\prime \prime}$ slit, providing a resolving power $\lambda / \delta \lambda \simeq 1100$. In photometric weather conditions, an additional low resolution HK band spectrum was taken through a $10^{\prime \prime}$ wide slit and used to recover the absolute flux calibration.

 

 

Table 1: Orders of the CASPIR grisms

grism	order	$\lambda_{\mbox{\tiny min}}$ (Å)	$\lambda_{\mbox{\tiny max}}$ (Å)
IJ	16	9700	10850
	15	10 290	11 490
	14	10 950	12 220
	13	11 700	13 050
	12	12 570	14 010
JH	17	12 470	13 890
	16	13 150	14 640
	15	13 920	15 480
	14	14 780	16 420
	13	15 770	17 490
HK	13	16 100	17 980
	12	17 320	19 320
	11	18 740	20 880
	10	20 400	22 710
	9	22 420	24 900

For optimal dark current and sky subtraction, "ABBA" observing sequences were followed. The star was shifted + or $- 4^{\prime \prime}$ along the $1^{\prime \prime}$ slit, or moved in and out of the $10^{\prime \prime}$ slit between A and B exposures. An ABBA sequence consisted of at least four 30 s exposures, each of which consisted of 2 to 100 individual readouts of the array depending on the brightness of the object. For the removal of telluric absorption features and the flux calibration, both O, B stars and late F or early G dwarf stars with known near-IR magnitudes were observed (typically one reference for every 4 or 5 programme observations).

2.2 Near-IR data reduction

The data was reduced with IRAF software, using the CASPIR reduction package (P. McGregor, CASPIR manual) and extensions thereof (A.L.). After subtraction of a median bias frame, a quadratic correction for the non-linearity of the response was applied (typically less than 1%) and ABBA sequences were combined. An internal incandescent lamp provided the light-on-light-off flatfields, which were then divided into the data frames. The traces of stars at two positions in the slit and both xenon and argon arcs were used to derive the geometrical transformation into straightened, wavelength calibrated images of the individual grism orders, from which the spectra were extracted and co-added. No additional sky subtraction was usually required. The wavelength calibration is accurate to better than 2 Å.

It is a well known problem that no spectrophotometric standards, in the sense familiar to optical observers, exist for near-IR flux calibration. The determination of the telescope, instrument and atmosphere transmission functions relies on the comparison of the observed reference star spectra with model atmosphere calculations. Ideally, objects with intrinsically featureless spectra are required for the removal of telluric features, and objects with known energy distributions for the relative spectrophotometric calibration. With 10 m class telescopes, increased sensitivities will make it possible to use featureless white dwarf spectra for the first purpose. Some currently feasible approaches based on A, F and early G type stars have been described by Hanson et al. (1996) and Maiolino et al. (1996). As far as telluric absorption is concerned the method followed here was similar, but extended to a much broader wavelength range. This spectral extent prohibits the use of blackbody curves to mimic the intrinsic energy distributions of the reference stars.

O, B and A stars have few intrinsic lines apart from the hydrogen (and sometimes helium) series and are easier to use than solar type references over most of the spectral range. However, their lines are often deep and broad, with complex profiles that vary from one line to the other, and the Brackett hydrogen lines are blended around 1.6 $\mu$m at our resolution. Removing them is not as easy as one could have hoped. In addition, early type stars are not evenly distributed on the sky and they are often found in dusty environments. Their use as relative flux standards requires the knowledge of the foreground extinction as well as their intrinsic energy distribution. F and early G dwarf stars display weaker but more numerous absorption features. Their energy distributions depart significantly from those of blackbodies, but at temperatures around 6000 K the Kurucz atlas (1993) provides reliable representations (as discussed in Lejeune et al. 1997; Castelli et al. 1997; Bessell et al. 1998). Our calibration procedure was based on both early and intermediate type reference stars and can be outlined as follows.

To select the most adequate model spectrum from the library of Kurucz (1993), the literature was searched reference star by reference star for individual estimates of the effective temperature, the surface gravity, the metallicity and the extinction on the line of sight (cf. Table 2). The corresponding Kurucz spectra were reddened when required, using the reddening law of Cardelli et al. (1989). All the raw extracted reference star spectra of a night were compared to each other and to theoretical spectra order by order. In this procedure, it was found extremely useful to be able to overlay early and intermediate type spectra in order to identify and distinguish between the strong stellar lines and the numerous telluric lines (cf. Fig. 2). The division of the models into the data, and interpolation across regions affected by strong stellar lines, then provided the combined throughput of the instrument and the atmosphere at the time of each reference observation.

Figure 2: Raw IJ, JH and HK spectra of the G0V star BS77 taken (from bottom to top) on one of the worst moist Summer nights (airmass 1.3), on a normal Summer night (1.6) and on a normal Winter night (1.2). The solar spectrum (Castelli 1996, private communication) is shown for comparison; its slope has been adjusted for display purposes. All spectra have been rescaled to a similar mean flux and then spaced. Note that changes in the slopes of the raw spectra are partly due to changes in the instrumental response between runs

Finally, each programme star was corrected with each of these transmission functions, the five orders of each grism were merged (there was no need for any additional scaling or shifting) and the results for each transmission function were compared. Depending on the stability of the weather conditions and of the instrument, the most important criterion for the selection of the final spectrum was either the similarity of the airmass or the closeness in time with the reference observation. The telluric CO₂ features around 2.00 and 2.05 $\mu$m are easy to identify and correlate tightly with airmass, whereas the broad H₂O absorption bands can vary independently of the airmass on timescales sometimes shorter than an hour. In about 20% of the cases, a second order correction of either the CO or the H₂O bands was performed, using ratios of transmission functions normalised to 1 in the atmospheric windows. However, since the recorded signal virtually drops to zero in the deepest water bands (especially in Summer), the energy distribution could only rarely be completely recovered.

A total of 28 stars were used as references over the course of the 21 near-IR observing nights, forming a network that allowed us to cross-check the model selections between nights. The reference stars and the model adopted for each of them are listed in Table 2.

When $10^{\prime \prime}$ slit HK spectra were available, they were corrected in a similar manner but an absolute flux calibration was also achieved, based on the known K band magnitudes of the reference stars. The high resolution spectra were then scaled to the level of the corresponding flux-calibrated low resolution ones.

 

 

Table 2: Reference stars for near-IR spectroscopy

	Star	Type	Nights (JD-2400000)	Model
1	BS33	F7V	17.6.95 (49886), 7.7.96 (50272)	t6250m00, Av=0.11
2	BS77	F9V	9.12.95 (50061), 30.1.96 (50113), 7.7.96 (50272)	t6000m00, Av=0.
3	BS674	B8IV-V	3.3.96 (50146)	t13000m00, Av=0.
4	BS818	F6V	20.11.97 (50773)	t6500m00, Av=0.03
5	BS1006	G2V	8.12.95 (50060), 7.7.96 (50272)	t5750m10, Av=0.
6	BS1291	F2V	8.12.95 (50060)	t6750m00, Av=0.05
7	BS1502	F2V	3.3.96 (50146), 7.7.96 (50272)	t7000m00, Av=0.
8	BS2015	A7V	8.12.95 (50060)	t7500m00, Av=0.13
9	BS2451	B8III	29.1.96 (50112), 20.11.97 (50773)	t11000m00, Av=0.04
10	BS3034	B0Ve	19.2.95 (49768)	t29000m00, Av=0.74, hot gas
11	BS3138	G0V	11.4.95 (49819), 8.12.95 (50060), 29.1.96 (50112)	t6000m10, Av=0.
12	BS3578	F7V	12.4.95 (49820), 8.12.95 (50060), 3.3.96 (50146)	t6000m10, Av=0.
			26.5.96 (50230), 27.5.96 (50231)
13	BS4102	F2IV	3.3.96 (50146), 30.3.96 (50173), 25.5.96 (50229)	t6750m00, Av=0.
			26.5.96 (50230), 8.7.96 (50273)
14	BS4133	B1Ib	16.2.95 (49765)	t20000m00, Av=0.21
15	BS4600	F6V	11.4.95 (49819), 25.5.96 (50229), 8.7.96 (50273)	t6500m00, Av=0.09
16	BS4638	B3V	16.2.95 (49765), 29.1.96 (50112), 30.1.96 (50113)	t18000m00, Av=0.15
17	BS4743	B2V	16.2.95 (49765), 8.7.96 (50273)	t20000m00, Av=0.12
18	BS4757	B9.5V	25.5.96 (50229)	t10000m00, Av=0.13
19	BS4773	B5V	18.6.96 (50253), 31.3.96 (50174), 26.5.96 (50230)	t15000m00, Av=0.05
			27.5.96 (50231)
20	BS4903	B1V	16.2.95 (49765)	t5750m00, Av=0.
21	BS4989	F7IV	3.3.96 (50146), 31.3.96 (50174)	t6000m05, Av=0.
22	BS5993	B1V	17.6.95 (50252), 26.5.96 (50230), 27.5.96 (50231)	t25000m00, Av=0.75
			6.7.96 (50271)
23	BS6310	F3V	2.3.96 (50145), 6.7.96 (50271), 8.7.96 (50273)	t6750m00, Av=0.
24	BS6314	F6V	11.4.96 (50185)	t6250m00, Av=0.
25	BS6496	F7V	18.6.95 (50253), 3.3.96 (50146)	t6250m00, Av=0.
26	BS7213	F7V	11.4.96 (50185)	t7000m00, Av=0.
27	BS7446	B1III	26.5.96 (50230), 27.5.96 (50231)	t25000m00, Av=0.88
28	BS7875	F8V	17.6.95 (50252), 7.7.96 (50272)	t6000m00, Av=0.03

Notes: Model descriptions are given as tXXXXmZZ, where XXXX is the effective temperature ( $T_{\rm eff}$), and ZZ indicates the metallicity ( $-\log(Z/Z_{\odot}$)). Theoretical spectra: Kurucz (1993), $\log (g)=4.5$ (using lower surface gravities for O, B stars has no significant effect on the near-IR colours). $T_{\rm eff}$ and metallicities were chosen according to the literature when available. Otherwise $Z_{\odot}$ was adopted and an approximate $T_{\rm eff}$ was derived from the spectral type; agreement with the known near-IR colours of the stars was given priority over compliance with a unique $T_{\rm eff}$- spectral type relation. Except for BS7446 (Theodossiou 1985), non-zero optical extinction values are from Neckel et al. (1980).

$\quad$ The model adopted for BS3034 combines the photosphere model with a gas emission model (Dachs & Wamsteker 1982). Both contributions are reddened and contribute equally to the emission at 2.45$\,\mu$m (in order that the resulting model colours agree with the observed ones to within 0.03 mag).

2.3 Optical spectra

Low resolution optical spectra were acquired quasi-simultaneously with the infrared spectra using the Reynolds Spectrograph on the 1.88 m Telescope at Mount Stromlo Observatory. The spectrograph was used with a 150 l/mm grating in first order (2 deg. Blaze angle, giving a dispersion of 9.7 Å/pixel), an order selecting filter with a cut-off at 5000Å (GG 495, but GG 550 in June95) and a 4 arcsec slit. The spectral resolution was usually seeing-limited, giving a FWHM resolution between 30 and 40Å instead of the slit-limited 50Å. Standard IRAF routines were used for the reduction of the data. The transformation to wavelengths coordinates was very close to linear. An additional zero point shift was required subsequently, correcting for the position of the star in the slit and instrumental instabilities: the wavelength of H$\alpha$ was forced to 6565 Å when present, and a cross-correlation in spectral regions with strong telluric features was used to determine the shift for the remaining spectra. The distribution of residual wavelength errors has a standard deviation of about 3 Å.

The flux calibration and removal of atmospheric features were based on a selection of spectrophotometric standards whose intrinsic spectra (themselves corrected for telluric absorption) were provided by M.Bessell.

   
2.4 Quality of the resulting spectra

The resulting optical data set consists of about 200 spectra with a typical signal-to-noise ratio per resolved element of 50, except in regions of strong telluric absorption (7605 Å, 9000 - 9800 Å). In the comparison of two optical spectra, changes in the (seeing limited) spectral resolution and calibration are the origin of most of the noise. The energy distribution between 5500 Å and 9000 Å is recovered to within 5% (1 $\sigma$), but the uncertainties rise to 10% below 5300 Å and around 9500 Å. V'-I (where V' is a V filter cut off below 5000 Å) can be measured to $\pm 0.15$ magnitudes and individual molecular features can be measured to within a few percent.

In the near-IR, there are 140 $0.97-2.49~\mu$m spectra of 55 programme stars (Table 3). The comparison between calibrations with various standards and the comparison of individual stars between nights indicate a typical signal-to-noise ratio per resolved element of 50 in atmospheric windows. It drops in the areas of strong telluric H₂O or CO₂ absorption shown in Fig. 2. On scales of $\sim$50 Å, bumps or dips at the 5% level sometimes remain as the residuals of an imperfect removal of internal reflections in the instrument (they are most obvious around 1.25 $\mu$m; however in most cases the features seen in this range are real).

 

 

Table 3: Programme stars: main properties, number of spectra

Name	1950	vartype	V	$\delta V$	P	epoch	Sp.type	Nobs	Figure
			(at min)		(days)	(at max)		(VK/VI/IK)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
Oxygen rich variables
WX Psc	010348 + 121951	M	>12.5	>2.5	645b	--	M9III(OH/IR)	0/0/1	B24
RS Hor	023436 - 624806	M	15.0	5.2	202.9	30100	M3e	3/3/3	B13
X Men	033340 - 763657	M	15.0	3.6	380.0	28720	M3e	4/4/4	B19
R Cnc	081348 + 115252	M	11.8	5.6	361.7	38093	M7IIIe	0/1/0
R Cha	082256 - 761136	M	14.2	6.7	334.6	42006	M4e-M7e	4/5/5	B10
SY Vel	091035 - 433416	SRb	10.2	1.3	63.0	--	M5/6III	2/3/2	B6
UZ Hya	091415 - 042348	M	14.5	5.7	261.0	42527	M4e	2/2/2	B16
CM Car	094644 - 671248	M	14.0	2.5	335.0	--	M2e	1/3/1	B11
S Car	100746 - 611814	M	9.9	5.4	149.5	42112	K5e-M6e	6/8/8	B7
KV Car	110108 - 665126	SRb	10.6	0.8	150.0	--	M4III	2/6/2	B4
RZ Car	103413 - 702729	M	15.0	6.2	272.8	39586	M4e-M8e	5/7/7	B8
RS Hya	104857 - 282143	M	14.4	5.2	338.6	39169	M6e	5/7/7	B15
U Crt	111015 - 070130	M	13.5	4.0	169.0	28269	M0e	2/5/3	B12
BD Hya	111342 - 295423	SRa	12.2	1.8	117.4	27900	M0e-M2	3/7/4	B14
SW Hya	130038 - 285014	M	13.0	3.2	218.8	29302	M2e	0/1/0
V603 Cen	130204 - 351922	M	13.4	2.3	253.0	38898	K0	0/1/0
T Cen	133853 - 332042	SRa	9.0	3.5	90.4	43242	K0e-M4e	4/5/4	B9
ET Vir	140806 - 160400	SRb	5.0	0.2	80.0	40697	M2III	1/2/1	B1
AFGL 1686	140839 - 073042	M	>12.5	>2.5b	700:b	--	M8III(OH/IR)	0/0/2.5a	B24
AO Cen	145734 - 421752	M	12.5	1.9	189.0	13627	M2e	0/2/0
S Lib	151831 - 201233	M	13.0	5.5	192.9	41883	M1.0e-M6.0e	2/3/2	B17
RS Lib	152124 - 224404	M	13.0	6.0	217.6	42154	M7e-8.5e	2/3/2	B18
R Cir	152357 - 573310	SRb	12.1	1.9	222.0	--	M4/6III	0/1/0
SV Lib	153020 - 270100	M	11.5	1.3	402.7	33852	M8e	3/5/3	B5
Z Lib	154336 - 205823	M	14.0	3.6	298.6	27689	M3e	0/1/0
R Ser	154823 + 151703	M	14.4	8.7	356.7	37010	M7IIIe	0/1/0
DX Ser	160600 - 012423	SRa	12.4	2.0	360.0	33880	M5-M8	0/1/0
WW Sco	162411 - 311133	M	14.7	4.1	431.0	19944	M6e-M9	2/3/3	B22
RY Cra	181818 - 445530	M	14.0	1.9	195.0	28845	M4e	1/1/2	B11
SV Tel	185228 - 493203	M	13.2	3.0	225.5	30045	M4e-M6e	1/1/2	B23
Z Aql	201231 - 061817	M	14.8	6.6	129.2	41938	M3e	0/0/2	B23
U Equ	205445 + 024710	M?	>15.5	>1	--	--	OH/IR	0/0/0.5a	B24
R Phe	235352 - 500355	M	14.4	6.9	269.3	39486	M2IIe-M4IIIe	4/5/5	B20
S Phe	235630 - 565114	SRb	10.6	2.0	141.0	--	M3e-M6IIIe	4/5/5	B21
Carbon rich variables
T Cae	044532 - 361750	SR	10.8	1.8	156	27840	C6,4(N4)	3/3/3	B27
R Lep	045720 - 145248	M	11.7	6.2	427.1	42506	C7,6e(N6e)	3/3/5	B31
RU Pup	080520 - 224600	SRb	12.2	1.9	425	--	C5,4(N3)	5/5/7	B30
Y Hya	094845 - 224657	SRb	12.0c	3.7c	302.8	--	C5,4(N3P)	6/7/7	B29
BH Cru	121337 - 560029	M	10.0	2.8	520d	40858	SC4.5/8e	3/5/3	B28
S Cen	122152 - 490947	SR	10.7	1.5	65	43242	C	1/2/1	B27
V Oph	162356 - 121854	M	11.6	4.3	297.2	45071	C5,2/7,4e(N3e)	0/1/0

Notes to Table3:
The columns are: Name (Col. 1), right ascension and declination (Col. 2), the variability type (Col. 3), V magnitude at minimum light (Col. 4), V amplitude (Col. 5), period (Col. 6), phase 0 epoch (visual maximum) expressed as JD-2400000 (Col. 7), and spectral type (Col. 8). All these quantities were taken from the GCVS (Kholopov et al. 1988). Column 9 gives the number of merged (VK), optical (VI) and infrared (IK) spectra available.
(a) A fractional value indicates an incomplete infrared spectrum. (b) Jones et al. (1990).
(c) AAVSO/AFOEV data suggest a significantly smaller amplitude, but non-detections need to be assessed for a quantitative statement (AFOEV 1999, Mattei 1999). (d) Period has increased (Bedding et al. 2000), GCVS period = 421d.

The uncertainties on the energy distribution (colours) result from the combined effects of instrumental changes, the focusing and positioning of the stellar image on the slit and the choice of a model for the intrinsic spectrum of the reference stars. Figure 3 illustrates the estimated quality of the relative flux calibration in the near-IR, again based on the comparison between reference star spectra taken at different times. After a systematic reduction of all data, we estimate that at least 85% of the spectra have errors smaller than indicated on the plot while the errors for the 15% most uncertain spectra don't exceed the values of Fig. 3 by more than 50%.

The 108 pairs of matching optical and near-IR spectra may be merged by scaling the spectra to the same mean in the most reliable part of the region of overlap (the range 9760 - 9870 Å). As mentioned above, this wavelength range suffers from relatively high uncertainties in the optical spectra, leading to an uncertainty of 0.15 mag in I-J and in all other colours combining optical and near-IR fluxes. Quadratic summing of the uncertainties across the optical and near-IR ranges, combined with the connection uncertainty at 9760 - 9870 Å, results in a total uncertainty of $\sim 0.3$ mag in $V^{\prime }-K$ (note that $V^{\prime}-K \sim 3.6-7.3$ for M 0-M 6 stars, respectively).

The potential user of the final data should not hesitate to contact the authors (A.L.) for additional information, especially if his/her aim is to study the detailed variations of one particular object or one particular spectral feature (instead of studying statistical properties).

 

 

Table 4: LMC/SMC stars (Fig.B26)

Star	1950	Period	K	J-K	Nobs
					(VK/VI/IK)
HV 2255 (LMC)	045808 - 701315	830	7.4	1.2	1/1/1
N371R20 (SMC)	005902 - 722647	580	8.2	1.0	1/1/1
HV 2446 (LMC)	052007 - 672308	602	9.2	1.2	1/1/1
HV 2360 (LMC)	051249 - 672308	790	7.7	1.1	1/1/1

 

 

Table 5: Bulge stars (Fig.B25)

Number	1950	Period	K	J-K	Nobs
NGC 6522 field:
205	175906 - 295917	470:	6.31	2.27	0/0/1
426	175928 - 300526	465	6.57	2.03	1/1/1
435	175917 - 300647	465	6.59	1.97	1/1/1
SGR I field:
4	175544 - 285032	450:	5.89	1.55	1/2/3
117	175735 - 291100	360	6.91	1.84	0/1/1
5	175512 - 284914	450:	6.47	1.78	0/1/0
21	175529 - 290113	355	7.96	1.64	1/2/2
11	175646 - 285323	750a	6.25	1.81	1/2/2
55	175618 - 290311	330	7.37	1.8	0/1/1

Notes to Tables 4 and 5:
The Bulge and Magellanic Cloud data have a lower signal-to-noise ratio than the local stars.
(a) Lloyd Evans (1976) finds P= 405 days.

 

Table 6: Red giant and supergiant stars

Name	2000		Type	Nobs	Figure	Comments
				(VK/VI/IK)
Giants:
BS 4104	102709	- 310404	K4III	0/1/0
BS 4432	113018	- 030012	K4.5III	1/2/1	B1
BS 5797	153803	- 423403	M0III	0/1/0
BS 4371	111443	+ 021708	M0III	1/2/1	B1
BS 2608	065616	- 484316	M1III	0/1/0
BS 4517	114551	+ 063146	M1III	0/1/0
BS 3923	095453	- 190046	M1III	1/2/1	B1
BS 4463	113513	- 472222	M3III	0/2/0
BS 4807	123822	+ 015117	M3III	0/1/0
BS 1695	050734	- 632359	M4III	0/1/0
BS 5603	150404	- 251655	M4III	0/1/0
BS 4532	114845	- 264524	M4III	0/1/0
BS 4267	105601	+ 061111	M5III	1/2/1	B1	$\delta V=0.3$
Supergiants:
HD 106873	1217 46	- 633658	M0I	1/1/1	B2	Lb, $\delta V=1.5$
HD 98817	1121 39	- 605928	M1I	1/1/1	B2
BS 3364	83030	- 364317	M2I	1/1/1	B2
HD 101712	114149	- 632452	M2-3I	2/3/2	B2
IRC -20427	180535	- 211339	M2I	0/0/1	B3
HD 115283	131725	- 613502	M4I	1/1/1	B2	$\delta V=0.6$
UY Sct	182736	- 122759	M4I	1/1/1	B3	SR, $\delta V=1.5$
V774 Sgr	175426	- 231409	M4-5I	1/1/1	B3	Lb, $\delta V=1.4$
EV Car	102022	- 602716	M4.5Ia	1/3/1	B3	SR, $\delta V=1.4$, P=347
CL Car	105400	- 610531	M5Iab	1/1/1	B3	SRc, $\delta V=2.50$, P=513

2.5 The final data

The reduced spectra are presented as a series of figures at the end of this paper. Merged optical/near-IR pairs are plotted when available, single near-IR spectra are plotted as next preference while single optical spectra are only shown in particular cases. The plotted spectra are normalised to a common mean flux per unit wavelength and then shifted for display purposes.

Figure 3: Uncertainty on broad band colour indices measured on CASPIR spectra of this programme, as a function of the separation between the central wavelengths of the two passbands defining the colour (avoiding the regions of heavy telluric absorption). Estimates are based on non-variable reference stars observed several times, and on differences between spectra obtained with the different flux calibration standards of one night. The solid line is an interpolation through the uncertainty estimates obtained by comparing fluxes in filters centered at 1.04, 1.25 and 1.65 $\mu$m, to the flux around 2.16 $\mu$m

All spectra are also available in numerical form through CDS. We provide the optical and near-IR spectra separately with names of the form nameVI.Rdate (ex. ScarVI.R1mar96) and nameIK.Cdate (ex. ScarIK.C3mar98), respectively, and, when applicable, the merged spectra of matching pairs with names of the form nameVK.date (ex. ScarVK.mar96). There are 182 VI spectra, 142 IK spectra and 108 VK spectra.

Finally, a series of tables made available electronically contain, for each spectrum:

-

the date of the observation;

-

the reference star used for removal of telluric features and flux calibration (IK spectra only);

-

the phase of the observation (when period and zero point available);

-

a flag indicating whether or not absolute flux calibration has been achieved;

-

an estimate of the stellar luminosity based on the period-K luminosity relation of Hughes & Wood (1990) and on the bolometric correction estimated from the spectrum itself (for merged VK spectra only).

Figure 4: Main band identifications for O-rich late-type stars. Left: a cool and a warm Mira spectrum below 1$\mu$m. Right: near-IR spectra; from top to bottom: the relatively warm Mira SCar, the cool Mira RCha and the red supergiant HD101712. Wavelengths are in Å, relative flux units are arbitrary