Sejong Open Cluster Survey (SOS) – II. IC 1848 cluster in the H ii region W5 West

Abstract

INTRODUCTION

There are three well-known large-scale star-forming regions (W3/W4/W5; Westerhout 1958) in the Cas OB6 association. W5 (also called the IC 1848 H ii region) is the eastern part of the molecular cloud complex and consists of two components W5 West and W5 East (Karr & Martin 2003). There are four known O stars, HD 17505, HD 17520, HD 237019 and BD +60 586 in W5 West, while one visible O star HD 18326 is at the centre of W5 East. All but HD 237019 are surrounded by many low-mass stars in clusters. These O stars are thought to be the ionizing sources of the H ii region. The simple morphology of each H ii region helps us to study the influence of strong ultraviolet (UV) radiation from massive stars on the surrounding materials and the triggering mechanisms for star formation (Koenig et al. 2008b). Hence, W5 is an ideal laboratory to study star formation processes with feedback from high-mass stars. The initial mass function (IMF) of this star-forming region is also an interesting scientific issue. In this context, our target of interest is the IC 1848 cluster (hereafter IC 1848), also known as OCL 364, situated in the centre of W5 West, containing the brightest star HD 17505, which is a multiple system consisting of an O6.5III, two O7.5V and an O8.5V stars (Hillwig et al. 2006).

Many investigators (Lada et al. 1978; Vallée, Hughes & Viner 1979; Normandeau, Taylor & Dewdney 1997 and therein) provided radio maps at various frequencies to investigate the large-scale structures and the physical properties of the W3/W4/W5 complex. Since active star formation takes place in these regions, investigators have a continuing interest in the processes involved. Loren & Wootten (1978) and Thronson et al. (1980) studied star formation in W5 A (or IC 1848 A), which is a bright-rimmed molecular cloud to the eastern end of W5. Wilking et al. (1984) discussed many interesting observational aspects associated with the triggered star formation in W5. A census of embedded stellar sources in the W3/W4/W5 region has been published by Carpenter, Heyer & Snell (2000). The authors identified 19 clusters, with about half the stars in five rich clusters. Karr & Martin (2003) identified several young stellar objects (YSOs) using the IRAS Point Source Catalog and found that the young stars are mainly distributed around the edge of the H ii region. The authors argued from a comparison of evolutionary time-scales that there is a distinct difference in the star formation history of the ionizing stars and the young stars on the border of expanding H ii regions. Because of the presence of clumps evaporating away from the ionizing sources and the shorter time-scale for star formation under the assumed expanding velocities, they also suggested that radiatively driven implosion – star formation triggered by the strong UV radiation from young OB stars (Elmegreen & Lada 1977) – may be the dominant triggering mechanism in W5. Spitzer/Infrared Array Camera (IRAC) and Multiband Imaging Photometer for Spitzer (MIPS) imaging data, Koenig et al. (2008b) provided photometric data for more than 17 000 point sources and YSO classifications. The identified YSOs made it possible to study the clustering properties of these YSOs and to discuss plausible triggering mechanisms required to explain the distinct generations within W5. The high-resolution IRAC and MIPS images revealed several isolated cometary globules and elephant trunk structures with polycyclic aromatic hydrocarbon (PAH) emission. Later Koenig & Allen (2011) studied the disc evolution of intermediate-mass stars. Cometary globules, bright-rimmed clouds and H ii region outflows are suggested as evidence of triggered star formation, and many studies have investigated the physical properties of these objects (Lefloch, Lazareff & Castets 1997; Thompson et al. 2004; Koenig et al. 2008a; Niwa et al. 2009).

Several photometric studies (Sharpless 1955; Hoag et al. 1961; Johnson et al. 1961; Becker & Fenkart 1971; Moffat 1972; Loktin, Gerasimenko & Malysheva 2001) involving IC 1848 have been made in the optical bands. These studies provided useful photometric data and fundamental parameters, such as reddening, distance and age for IC 1848. However, most analyses assumed a normal reddening law (R_V ∼ 3.0), and the multiplicity of early-type stars was not fully taken into account. Furthermore, the limiting magnitudes were not deep enough to study the pre-main-sequence (PMS) stars fainter than V ∼ 16 mag in the young clusters or associations in the Perseus arm.

Recently, a deep UBVI_C photometric study for IC 1848 W5 East has been made by Chauhan et al. (2011) with archival data. The authors presented the initial mass function (IMF) for stars in the mass range from 0.4 to 30 M_⊙, as well as their fundamental parameters. However, no deep optical photometric study has been made for IC 1848–W5 West. Since the results of Koenig et al. (2008b) provided very good membership criteria for the PMS stars with accretion discs, the age distribution and mass accretion rates of PMS stars, the star formation history within the region and the IMF down to the 1.0–1.5 M_⊙ regime can be studied. In addition, the Two Micron All Sky Survey (2MASS; Skrutskie et al. 2006) and Spitzer/IRAC data (Koenig et al. 2008b) make it possible to test carefully the reddening law in a wide wavelength range.

The previously published distance to IC 1848, derived using various methods, is in the range 1.7–2.4 kpc. The photometric surveys mentioned above provided distances of 1.7 kpc (Sharpless 1955), 2.1 ± 0.3 kpc (Chauhan et al. 2011), 2.2 ± 0.2 kpc (Johnson et al. 1961; Loktin et al. 2001) and 2.3 kpc (Becker & Fenkart 1971; Moffat 1972). Some studies have implicitly assumed that the distance to IC 1848 is the same as that to W3 or W4 (or IC 1805) (1.9–2.4 kpc). Georgelin & Georgelin (1976) independently derived a distance of 2.3 kpc using their Galactic rotation model. Early studies have relied on this distance. Recently, Xu et al. (2006) derived a distance of 1.95 ± 0.04 kpc for a W3OH maser using its very long baseline interferometry (VLBI) parallax. Hillwig et al. (2006) preferred to use a distance of 1.89 kpc to match the observed and theoretical stellar radii. Since distance is a critical parameter in converting observational quantities to reliable physical parameters, we need to revisit its determination.

The Sejong Open Cluster Survey (SOS) project is dedicated to provide homogeneous photometric data down to V ∼ 22 mag for many open clusters. The overview of the project can be found in Sung et al. (2013b, hereafter Paper 0). The young open cluster NGC 2353 was studied as part of this project (Lim et al. 2011). This paper on IC 1848 is the third in the series. The observations and comparisons with previous photometry are described in Section 2. In Section 3, we discuss the reddening law in the direction of IC 1848 and present fundamental parameters estimated from the photometric diagrams. The IMF of IC 1848 and the mass accretion rates of PMS stars with UV excesses are presented in Sections 4 and 5, respectively. Finally, we summarize the results from this study in Section 6.

OBSERVATION

The observations of two regions associated with IC 1848 were made on 2009 January 19 using the AZT-22 1.5-m telescope (f/7.74) at Maidanak Astronomical Observatory in Uzbekistan. All imaging data were acquired using the Fairchild 4098 × 4098 CCD (SNUCam; Im et al. 2010) with standard Bessell UBVI (Bessell 1990) and Hα filters. Lim et al. (2008) have described the characteristics of the CCD in detail. The mean seeing was better than 1.0 arcsec, and sky conditions were good. The observations comprised a total of 20 frames that were taken in two sets of exposure times for each band – 5 and 60 s in I, 5 and 180 s in V, 7 and 300 s in B, 15 and 600 s in U and 30 and 600 s in Hα. We present the finder chart for the stars brighter than V = 15 mag in Fig. 1 using the Guide Star Catalog version 2.3 (Lasker et al. 2008). The photometry for 12 stars that were saturated in our data was taken from previous studies. The photometric data and the adopted spectral types of the stars are presented in Table 1.

Figure 1.

Finder chart for IC 1848. The size of the circles is proportional to the brightness of individual stars. The position of stars is relative to the brightest O-type star HD 17505 (⁠|$\alpha = 02^{\rm h} 51^{\rm m} 08 {.\!\!\!\!^{{\mathrm{s}}}}0$| δ = +60°25^′04^′′, J2000). Squares outline the observed regions.

Open in new tabDownload slide

We checked our photometry against previous studies in order to confirm its consistency. Only a few photoelectric photometric data sets for IC 1848 were found in the open cluster data base WEBDA.² In addition, no photometric study with a modern CCD was reported for IC 1848. The photoelectric photometric data of Hoag et al. (1961) were compared with our CCD data for 9–11 common stars. The differences in the V magnitude, the B − V and U − B colour indices are, respectively, ΔV = 0.021 ± 0.020 mag, Δ(B − V) = 0.001 ± 0.054 mag and Δ(U − B) = 0.008 ± 0.067 mag, where Δ ≡ this − Hoag et al. Although the scatter seems to be rather large, our photometric zero-points are in good agreement with those of Hoag et al. (1961). Since the photoelectric data of Hoag et al. (1961) are well consistent with those of Johnson & Hiltner (1956), we can conclude that our photometry is well tied to the Johnson standard UBV system. On the other hand, it is impossible to directly test the consistency of our V − I colours due to the absence of photoelectric or CCD photometric data. However, we were able to compare our CCD photometric data for the young open cluster NGC 1893, which was observed on the same night as IC 1848, with the CCD photometric data of other studies and found that our data are well consistent with previous studies (Massey, Johnson & DeGioia-Eastwood 1995; Sharma et al. 2007). In the comparisons with the photometric data of other studies no evidence for any difference depending on colour indices was found. The details of the comparisons will be presented in the forthcoming paper (Lim et al., in preparation). In addition, as the ratio of total-to-selective extinction obtained from the colour excess ratio of E(V − I)/E(B − V) is consistent with that from other colour indices (see Fig. 6), our V − I colour is well tied to the Cousins standard system.

PHOTOMETRIC DIAGRAMS

The fundamental parameters of open clusters are very important to study many scientific issues, such as the local spiral arm structure of the Galaxy, observational test of stellar evolution theory, the stellar IMF, the history of the star formation and chemical evolution in the Galactic disc, as mentioned in Paper 0. Such parameters should be carefully determined by using homogeneous photometric data. Traditionally, fundamental parameters of open clusters are determined from optical photometric diagrams. Although near-infrared (NIR) photometry gives much insight into faint or obscured objects owing to its less sensitivity to interstellar reddening, optical photometry is still a powerful tool to derive many fundamental parameters of stars, because it has well-calibrated empirical relations and provides relatively higher resolution colour indices for massive stars. Hα photometry (Sung, Bessell & Lee 1997) and the classifications of YSOs in the W5 region (Koenig et al. 2008b) provide very useful membership constraints for the PMS stars in IC 1848. In this section we present the membership selection criteria, the reddening law and fundamental parameters of IC 1848 from the two-colour diagrams (TCD) in Fig. 2 and the colour–magnitude diagrams (CMD) in Fig. 3.

Figure 2.

Colour–colour diagrams of IC 1848. In the left-hand panel, bold dots, triangles (red), squares (blue), crosses (blue), open circles (red) and pentagons represent early-type members, Class I, Class II, transitional disc candidates, Hα emission stars and bright stars obtained from previous studies, respectively. The intrinsic and reddened colour-colour relations are overplotted with a solid and dashed line, respectively. The mean reddening of 〈E(B − V)〉 = 0.66 mag is adopted for the latter. In the right-hand panel, the solid line represents the empirical photospheric level, while the dotted and dashed lines are the lower limits of Hα emission candidates and Hα emission stars. From this criteria, 196 Hα emission stars and 35 candidates are identified.

Open in new tabDownload slide

Figure 3.

CMD of IC 1848. Left-hand panel: V − I versus V diagram. Middle panel: B − V versus V diagram. Right-hand panel: U − B versus V diagram. The solid lines represent the reddened ZAMS relations of Paper 0 shifted by E(B − V) = 0.66 mag and V₀ − M_V = 11.7 mag. The arrow denotes a reddening vector corresponding to A_V = 3 mag. The other symbols are the same as Fig. 2.

Open in new tabDownload slide

Membership selection

Early-type members (from O to late-B) can be selected from their photometric properties in the CMDs and the (U − B, B − V) TCD. The criteria for the early-type members are (1) V ≤ 15, 0.2 ≤ B − V ≤ 0.7, −1.0 ≤ U − B ≤ 0.5 (see left-hand panel of Figs 2 and 3), E(B − V) ≥ 0.5 and −1.0 ≤ Johnson's Q ≤−0.1; (2) an individual distance modulus between |$\langle V_0 - M_V \rangle _{\mathrm{cl}} - 0.75 - 2.5\sigma _{V_0 - M_V}$| and |$\langle V_0 - M_V \rangle _{\mathrm{cl}} + \ 2.5\sigma _{V_0 - M_V}$| to take into account the effect of binary members and photometric errors (Sung & Bessell 1999; Kook, Sung & Bessell 2010; Lim et al. 2011), where 〈V₀ − M_V〉_cl and |$\sigma _{V_0 - M_V}$| are the mean distance modulus and the width of the Gaussian fit of the distance modulus, respectively. The reddening of individual members was determined and corrected for (see Section 3.2). The distance to the star was estimated using the zero-age main-sequence (ZAMS) relation from table 3 in Paper 0. Finally, we compared it with the average distance and the standard deviation of all members and candidates. In this procedure one star (ID 12815) was rejected from the list of members. Another star (ID 6138) was also rejected from the [E(V − I), E(B − V)] colour excess ratio (Fig. 6) due to a large deviation from the global trend. The star may be a foreground late-type star. We repeated this procedure until no further stars were rejected. In the end, 105 early-type stars were selected as members of IC 1848.

Since many PMS stars with an accretion disc exhibit X-ray emission, UV excess, Hα emission and an infrared excess, Hα photometry is an efficient way to select PMS members for young open clusters (≤3 Myr). Indeed, Sung et al. (1997) have successfully identified PMS members in NGC 2264 using Hα photometry. The authors and their collaborators have used Hα photometry to select PMS members in many young open clusters, NGC 6231 (Sung, Bessell & Lee 1998; Sung, Sana & Bessell 2013a), NGC 6530 (Sung, Chun & Bessell 2000), NGC 2244 (Park & Sung 2002), NGC 2264 (Park et al. 2000; Sung et al. 2008), NGC 3603 (Sung & Bessell 2004) and Trumpler 14 and 16 in the η Carina nebula (Hur, Sung & Bessell 2012). We found 196 Hα emission stars and 35 candidates using the Hα index defined in the previous studies (Sung et al. 2000) and presented the stars (open circle) in the right-hand panel of Fig. 2. Stars which exhibit a weak Hα emission line are likely missed in a single observation because the strength of the Hα emission line varies with time. We included 26 additional stars with emission equivalent widths of Hα larger than 5 Å from Koenig & Allen (2011), which were not identified as Hα emission stars in our photometry.

Excess emission by dust in the circumstellar disc can be detected at NIR and mid-infrared (MIR) wavelengths. The excess due to dust emission is more prominent in the MIR than the NIR. But prior to Spitzer, the Wide-Field Infrared Survey Explorer (WISE) and Herschel era, ground-based observations at NIR wavelengths were restricted to large IR excess stars. As a result, many studies have attempted to identify PMS members through strong NIR emission in the (J − H, H − K) TCD. Now, owing to the vast survey of the W5 region in the MIR (Koenig et al. 2008b), we could easily select the young stars with a circumstellar disc. The optical counterparts of the MIR excess emission stars in Koenig et al. (2008b) were searched within a searching radius of 1.0 arcsec, and we identified 397 YSOs (5 Class I, 368 Class II, 24 transition disc candidates). A total 462 PMS members were identified in our photometry.

As seen in the (V, V − I) CMD of Fig. 3 or Fig. 8, our membership selection constrains the PMS locus well. However, current membership are certainly incomplete, because the membership criteria of PMS stars mentioned above are biased towards stars with an active accretion disc. However, it is worth comparing the detection efficiency of low-mass PMS stars between Hα photometry and MIR data for stars brighter than V = 19 (our photometric completeness is about or better than 88 per cent). A total of 94 stars were classified as low-mass PMS members either from MIR excess emission or Hα photometry. Among them, 32 stars were detected from both MIR and Hα photometry. Interestingly, none of five bright transitional disc candidates (V ≤ 16) shows Hα emission. However, the Hα detection fraction of faint transitional disc candidates (3/8 = 38 ± 22 per cent to V = 16–19 mag) is very similar to that of Class II objects (26/77 = 34 ± 7 per cent). This fact implies that the faint transitional discs have similar characteristics to Class II objects, i.e. they are probably pre-transition disc objects (Sung, Stauffer & Bessell 2009). The reason for the absence of Hα emission among bright PMS stars with a transitional disc may be related either to the absence of a hot inner disc or to weak Hα emission with Hα emission equivalent widths smaller than 10 Å (Sung et al. 2008), or both. The detection efficiency of transitional disc candidates (V ≤ 19 mag) from Hα photometry may be less than 23 per cent with respect to MIR excess emission. In this work, MIR excess appears to be a more efficient way to find PMS members than Hα photometry.

However, the detection efficiency from MIR and Hα photometry is not as high as X-ray observation and strongly depends on the age of a cluster because the excess emission is prominent in young PMS stars with a circumstellar disc and accretion activity. In addition, we know that weak-line T Tauri stars (WTTS) are unlikely to be selected as PMS members from MIR or Hα photometry because most of the material in the disc of such stars has dissipated. Indeed, in NGC 2264, many WTTS without Hα emission or a MIR excess were identified as PMS by X-ray observations (Sung et al. 2009), and their photometric properties were found to be similar to those of MS stars (Flaccomio et al. 1999). Almost complete lists of PMS members in NGC 2264 and NGC 6231 have now been identified down to their limiting magnitude using X-ray sources from the Chandra data archive (Sung et al. 2008) or from XMM–Newton observation (Sana et al. 2006). Up to now, no extensive X-ray observations for the IC 1848 region exist, so we note that a large number of WTTS or low-mass PMS stars with weak accretion activity are likely missed in our membership selection.

The reddening

The interstellar reddening towards young open clusters is, in general, derived by comparing the observed colour indices of the early-type members with the intrinsic (U − B, B − V) diagram. Using the intrinsic colour–colour relation from table 1 in Paper 0 we determined the individual reddening E(B − V) of early-type members of IC 1848. The reddening slope was simply assumed to be E(U − B)/E(B − V) = 0.72 because the dependence on E(B − V) is negligible for less reddened stars [E(B − V) < 1 mag]. The mean reddening value was estimated to be 〈E(B − V)〉 = 0.660 ± 0.054 (s.d.) mag and displayed in the left-hand panel of Fig. 2 as a dashed line. This value is in good agreement with that of previous studies, e.g. E(B − V) = 0.55–0.70 mag (Johnson et al. 1961), 0.66 mag (Becker & Fenkart 1971), 0.72 mag (Moffat 1972) and 0.60 mag (Loktin et al. 2001).

From the spread of the early-type members in the (U − B, B − V) diagram, one can see that there is a non-negligible amount of differential reddening across the observed region. In order to investigate the spatial variation of the reddening we constructed the reddening map of the observed regions using the position and reddening of individual early-type members. The observed field was divided into 10 000 rectangular areas, each of which has the size of 0.33 × 0.23 arcmin² in right ascension and declination, respectively. The weighted-mean reddening of each point is calculated from the reddening of 105 early-type members, where the weight is exponentially decreasing with the distance from individual early-type members. We present the reddening map of IC 1848 in Fig. 4. The massive molecular cloud W5 NW in W5 West lies in the northern part of the cluster (see fig. 1 of Wilking et al. 1984; Koenig et al. 2008b). O-type stars, HD 17505, HD 17520 and HD 237007, in W5 West are responsible for the ionization of the molecular clouds (Karr & Martin 2003), and bright rims lie along the ionization front (Wilking et al. 1984). These could be evidence that the stellar wind and UV radiation from three O-type stars are compressing the molecular cloud. On the other hand, the reddening map of the southern part is rather transparent compared to the northern part, and one could expect that the material in the southern part might have been rapidly swept away. Thus, the large-scale variation of reddening shows the systematic difference between the northern and southern part of the cluster. Another O star BD +60 586 (Δα ∼ 22.5 arcmin, Δδ ∼ 14.0 arcmin) is located in the valley of reddening between two molecular clouds W5 NE and W5 NW. The structure in the reddening map may reflect the interaction between the O stars and the surrounding materials. A similar feature was found in NGC 6231 (Sung et al. 2013a). From this, we deduce that the strong stellar wind from O stars can sweep away the surrounding medium. In contrast, Lim et al. (2013) found the highest reddening value to be in the centre of the starburst cluster Westerlund 1 and interpreted this as resulting from dust formed from the ejected material from the many Wolf–Rayet stars. These observations indicate that within IC 1848, dust destruction may be more dominant than dust production. However, unlike the other O-type stars, HD 237019 (Δα ∼ 17.3 arcmin, Δδ ∼ 2.5 arcmin) lies in a highly reddened region. According to a formation scenario of W5 West suggested by Koenig et al. (2008b) the star could be ejected from the cluster by gravitational interaction between massive stars in the cluster centre. If this scenario is true, the star may not have had enough time to interact with the surrounding material.

Figure 4.

Reddening map of IC 1848. Squares outline the observed regions. The size of the filled circles is proportional to the brightness of individual early-type stars. Reddening contour lines are shown for E(B − V) = 0.60 (dashed line), 0.65 (dotted line) and 0.70 mag (solid line). Large open circles indicate the O stars.

Open in new tabDownload slide

To check the reliability of the reddening corrections for PMS stars in our analysis, we compared the reddening E(V − I) from Fig. 4 with that from the spectral type. Recently Koenig & Allen (2011) published the spectral types of 389 stars in the W5 region. Among them, 63 stars with V − I, B − V and U − B colour indices in our data were used in this comparison. The spectroscopic reddening was obtained by comparing the observed V − I colour with the intrinsic colour inferred from a given spectral type (Paper 0). One can expect that the mean difference would be close to zero with a random scatter, and that spectroscopic reddening could be higher for few stars due to their internal extinction. We present the comparison in the left-hand panel of Fig. 5, where negative values mean that spectroscopic reddening is larger than that from our reddening map. For B–early-F and K5–late-K-type stars, the reddening shows reasonably good agreement. However, the spectroscopic reddening is 0.2 mag higher than ours for late-F–K4-type stars, and systematically lower for M-type stars.

Figure 5.

Differences between the reddening adopted from the reddening map and reddening derived from the spectral-type intrinsic V − I colour relation plotted against spectral type and UV excess. Δ means the reddening from Fig. 4 minus the spectroscopic reddening. Circles, open squares, triangles, filled squares and diamonds represent B0–G0, G1–K0, K1–K4, K5–K9 and M-type stars, respectively. Δ(U − V) means UV excess derived from this work. The stars with negative Δ(U − V) values are UV excess stars. See the main text for details.

Open in new tabDownload slide

In order to find the main causes of the differences we checked our intrinsic relations in Paper 0, which is based on the (B − V, V − I) relation by Cousins (1978) with the spectral type–colour relation of Bessell & Brett (1988). The difference is less than 0.02 mag, and therefore it is unlikely to be caused by the spectral type–(V − I) relation. We also compared the reddening map from optical photometry with that from the 2MASS data. With E(V − H) and E(V − K_S) of the early-type members (see next section) the NIR reddening map was constructed through the same procedure as the optical reddening map. The spatial variation of the E(H − K_S) reddening is compatible with that of the optical photometry. In addition, we also checked the spatial distribution of the G–M-type stars which showed a systematic difference in reddening derived from the two different methods. If our reddening map exclusively provides small/large value of reddening towards any specific direction, this method of reddening correction may be inappropriate for the PMS members. However, the late-type PMS stars were distributed uniformly, thus the application of the reddening map may have nothing to do with the discrepancy.

The other possibility is that the spectral types of the stars were classified as earlier types due to veiling filling in some of the lines. Koenig & Allen (2011) described their classification schemes without any discussion on the veiling effect. Since the software SPTclass (Hernández et al. 2004) provides the spectral type based on the equivalent width of the spectral lines of O8–M6-type MS stars (see their webpage),³ veiling effect could make the spectral classification of PMS stars tricky. The classification schemes in SPTclass are optimized for Herbig Ae/Be, late-F–mid-K and mid-K–M-type stars, respectively. The classification for late-F–mid-K-type stars includes 11 spectral indices in the relatively narrow spectral region (from 4200 to 6200 Å), and thus it can be affected by veiling. Furthermore, there are fewer number of spectral indices than for Herbig Ae/Be and later type stars. Hence, the spectral classification of PMS stars could be misled by the presence of non-photospheric contributions. According to the spectral indices for F, G and K-type stars, e.g. Ca i λ4226, Fe i λ4458 and Fe i λ5329, the equivalent width of the lines increases with spectral type. If optical spectra of PMS stars with veiling exhibit decreased line depths, the equivalent widths of the spectral indices will appear to be that of earlier types. Therefore veiling would shift the spectral classification to earlier types. If the speculation is correct, the stars should exhibit a UV excess. We plot UV excess, Δ(U − V), derived in Section 5, with respect to the discrepancy in the reddening E(V − I) in the right-hand panel of Fig. 5. The stars possibly classified as earlier spectral type are expected to be located in the lower left-hand side of the panel. Indeed, they (F0–K4-type PMS stars – circle, open square and triangle) locate in the expected region, while stars without a UV excess are clustered at ΔE(V − I) = 0 and Δ(U − V) = 0.

However, K5–M-type stars show a systematic and opposite trend. As the strength of TiO and CaH bands are used for the classification of mid-K to M-type stars, the spectral classification should be correct. In addition, these red bands would be little affected by non-photospheric contributions. As shown in the right-hand panel of Fig. 5, K5–K9-type stars (filled square) are close to zero in ΔE(V − I), although most of them exhibit a UV excess. However, three (ID 10938, 13480 and 14932) out of four M-type stars with V − I and U − V colour indices show a small spectroscopic reddening E(V − I)_sp of ∼0.2 mag, which is an unrealistic value for IC 1848. We also checked the position of the stars in Fig. 8 and found them to lie in the PMS locus. These facts imply that they have too late a spectral type. According to Koenig & Allen (2011) those stars are identified as Class II objects with strong Hα emission, and our Hα photometry also tags them as Hα emission stars. They must be very young stars of IC 1848, and therefore we attribute this discrepancy to a gravity effect on the spectra of young PMS stars. Many studies (Luhman, Liebert & Rieke 1997; Briceño et al. 1998; Luhman et al. 1998; White et al. 1999) remark that the spectral features of young PMS stars between 6000 and 9000 Å resemble those of giant stars and we note that the inclusion in the spectral classification methodology of a Na i λ8183/8195 index, that is very gravity sensitive, would help discriminate between giant- and dwarf-like M spectra. Given the weaker TiO bands of MS stars compared to giant stars, it is possible that the stronger bands of a PMS star can lead to them being assigned a later dwarf spectral type.

The same trend as in Fig. 5 can also be found in the published data (A_V and spectral type) of Koenig & Allen (2011). We investigated their A_V values with respect to the spectral types. The A_V values of G–K-type stars appear to be higher than those of other spectral types. The mean A_V of B–F0-type stars is about 1.5 mag, but the A_V values of G5-type stars have a range of 2.0–2.6 mag. The differences between two groups with different spectral types in E(V − I) is about 0.21–0.45 mag. These values are in good agreement with those found in Fig. 5. In addition, the A_V values of K–M-type stars decrease with the subclasses. The feature is the same as that found in Fig. 5. These facts imply that the discrepancy found in Fig. 5 originates from the published data of Koenig & Allen (2011). We therefore conclude that the application of our reddening map is more reliable than the reddening correction inferred from previously published spectral types in the current state and does not cause serious problems, except for a few PMS stars with nearly edge-on discs.