MULTI-WAVELENGTH STUDIES OF SPECTACULAR RAM-PRESSURE STRIPPING OF A GALAXY. II. STAR FORMATION IN THE TAIL

For a detailed analysis of the Hα counterpart along the NGC 4388 tail, we retrieved data obtained with the Subaru Prime Focus Camera (Suprime-Cam; Miyazaki et al. 2002) around the H i tail in the N-A-L659 filter (Yoshida et al. 2002; Okamura et al. 2002; Hayashino et al. 2003, NA659 hereafter) and the R (W-C-RC)-band filter. The center wavelength of NA659 is 6600 Å and the FWHM is 100 Å, which corresponds to Hα at a recessional velocity of v = 1700 ± 2300 km s⁻¹. The NA659 imaging dataset was the same as that used in Yoshida et al. (2002), but we reprocessed it from the raw data for this paper. Recently, the region was observed much deeper in the R-band. We used the data for the continuum subtraction. Also, V (W-J-V), i (W-S-I+), and N-A-L503 (Yoshida et al. 2002; Okamura et al. 2002; Hayashino et al. 2003, NA503 hereafter) band data were available. The center wavelength of NA503 is 5020 Å and the FWHM is 100 Å. NA503 covered [O iii]4959 and [O iii]5007 at a recessional velocity of v = 3680 ± 3020 and v = 780 ± 3000 km s⁻¹ and did not cover Hβ in the Virgo Cluster. The details of the Suprime-Cam data reduction are given in Appendix A.1. In Appendices A.2–A.6, the data in other wavelengths are described.

Based on a visual inspection, we selected two fields for spectroscopic observations. We call the four clumps of the Hα-emitting regions HaR-1, 2, 3, and 4, from the north to the south (Figure 1). Their appearances are shown in Figures 2 and 3 and their coordinates are given in Table 1. HaR-2 consists of several sub-clumps. We call them HaR-2a, 2b, 2c, and 2d, from the north to the south, respectively. Their multiband magnitudes are given in Tables 2 and 3. It should be noted that there are other Hα-emitting regions around the tail. The sampling of the Hα-emitting regions in this study was not complete.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

2.2.1. FOCAS Observations and Data Reduction

We carried out longslit spectroscopy in the two fields (HaR-1,2 and HaR-3,4) on 2013 March 3 (UT) with FOCAS (Kashikawa et al. 2002) attached to the Subaru Telescope. We used a longslit 0.8 arcsec wide, and the 300B grism without order-cutting filters. The dispersion was ∼1.38 Å pixel⁻¹. The pixel scale of the FOCAS data was 0.207 arcsec along the slit.

The data reduction was done in a standard manner: the overscan was subtracted and flat-fielded, cosmic rays were removed by a Python implementation of the LA Cosmic package (van Dokkum 2001),⁹ and the wavelength was calibrated using sky emissions and a thorium-argon comparison lamp obtained during the observations. The measured instrumental profile was ∼9 Å FWHM. Then, the sky emissions and the absorption features of the sky background (mainly moonlights) were subtracted. The two-dimensional (2D) spectra around Hβ and Hα are shown in Figure 4.

Zoom In Zoom Out Reset image size

Feige 67 was observed with a longslit 2.0 arcsec wide as a spectrophotometric standard. The calibrated spectrum was obtained from calspec in STScI.¹⁰ The atmospheric dispersion corrector at the Cassegrain focus was not available during the observations and a wavelength-dependent positional shift of the target across the slit should have occurred. In our observing conditions, differential atmospheric dispersion was negligible for HaR-3,4, while a ∼13% loss of flux may exist in the worst case for HaR-1,2. We therefore regard the Hβ/Hα ratio of HaR-1 and 2 as possibly having suffered a 13% systematic error.

2.2.2. Spatial Distribution and Apertures

The spatial distribution of flux around Hα, [N ii]6584, and the continuum is shown in Figure 5. The Hα and [N ii] fluxes are not only due to emission lines but also include continuum flux. In HaR-1 and 4, Hα, [N ii], and the continuum profiles are similar. In HaR-3, the continuum shows a small gradient, which could be contamination from a neighboring galaxy at the position x ∼ 21.

Zoom In Zoom Out Reset image size

In HaR-2, the spatial profile was rather complicated. At x ∼ 24, Hα and [N ii] show peaks while the continuum does not. At x ∼ 25, the continuum has a peak and Hα and [N ii] show a bump. Around x ∼ 27, Hα and [N ii] show a small peak, but the continuum is ∼0. Then, a second peak in the continuum is seen at the slit position x ∼ 29, where Hα and [N ii] are not so strong as in the other part. We therefore divide the aperture of HaR-2 into four parts, which correspond to the clumps in Figure 2 (bottom). Then, spectra were extracted for the Hα peaks with a Gaussian weight along the slit. The spectra are shown in Figure 6. In the figure, the wavelength dependence of the instrumental and atmospheric throughput is corrected, except for the possible slit loss. The neighboring galaxy between HaR-3 and 4 was found to be a background star-forming galaxy at z = 0.118.

Zoom In Zoom Out Reset image size

2.2.3. Measurement of the Line Strengths

The strength of Hα, [N ii]6584, [S ii]6717,6731, [O iii]5007, and Hβ were measured by Gaussian fitting. Over 6560 Å  <  λ  < 6820 Å, Hα, [N ii], and [S ii] were fit by a multivariate least-squares fitting. We assume that the background continuum is constant in the fitting region, that the redshift is the same for the lines, and that the FWHM of the lines is the same (dominated by the instrumental profile). The strength of [N ii]6548 was fixed to be 1/3 of that of [N ii]6584. The estimated redshifts are shown in Table 4. Over 4800 Å  <  λ  <  5100 Å, Hβ and [O iii]5007 are measured in the same way. [O iii]4959 was assumed to be 1/3 of [O iii]5007. At HaR-2c and HaR-2d, the lines are too weak and we could not fit well.

Table 4. Recessional Velocity in the Heliocentric Reference Frame and Projected Distance from NGC 4388

Name	z	v	d
		(km s−1)	(kpc)
HaR-1	0.00749	2.25 × 103	67.1
HaR-2a	0.00737	2.21 × 103	66.2
HaR-2b	0.00737	2.21 × 103	66.1
HaR-2c	0.00736	2.21 × 103	66.0
HaR-2d	0.00731	2.19 × 103	65.9
HaR-3	0.00780	2.34 × 103	35.9
HaR-4	0.00789	2.37 × 103	34.9

Download table as:  ASCIITypeset image

Since the seeing size was comparable to the slit width and the slit width was narrower than the spatial extent of the HaRs, the flux value from the spectra has little physical meaning but the flux ratios are meaningful. The ratios are shown in Table 5. The error of each parameter was estimated by a residual bootstrap method with 1000 realizations, except for Hα/Hβ. The error of Hα/Hβ was estimated from the error of the Hα flux and that of the Hβ flux by residual bootstrapping, assuming that their errors were independent. For HaR-1 and HaR-2, a 13% error due to a possible differential atmospheric dispersion effect was also included.

Table 5. Line Ratios

Name	[N ii]6584/Hα	[S ii]6717,6731/Hα	[S ii]6717/[S ii]6731	[O iii]5007/Hβ	Hα/Hβ
HaR-1	0.12 ± 0.02	0.09 ± 0.02	0.88 ± 0.49	3.79 ± 0.57	5.4 ± 0.8a
HaR-2a	0.22 ± 0.01	0.27 ± 0.02	1.45 ± 0.17	0.35 ± 0.05	4.0 ± 0.3a
HaR-2b	0.22 ± 0.01	0.28 ± 0.02	1.34 ± 0.17	0.23 ± 0.07	3.7 ± 0.4a
HaR-2c	0.22 ± 0.04	0.35 ± 0.05	2.39 ± 0.92	0.07 ± 0.25	6.1 ± 3.0a
HaR-2d	0.17 ± 0.06	0.37 ± 0.08	1.56 ± 0.67	... b	>4.4a,c
HaR-3	0.22 ± 0.01	0.19 ± 0.01	1.44 ± 0.14	0.93 ± 0.04	4.4 ± 0.2
HaR-4	0.25 ± 0.01	0.40 ± 0.02	1.29 ± 0.11	0.31 ± 0.07	3.7 ± 0.3

Notes. ^aIncluding 13% error due to possible differential atmospheric dispersion effects. ^b[O iii] and Hβ non-detection. ^cFrom the upper limit of Hβ.

Download table as:  ASCIITypeset image

In Figure 7, diagnostic line ratio plots are shown. All the regions have a line ratio similar to an H ii region. Moreover, [O i]6300 was not detected in any of the regions and a possible contribution of shock, as seen in Hα knots in Yoshida et al. (2004), are negligible in our targets. Note that [O i]6300 of the HaRs is not affected by the atmospheric [O i]6300 or [O i]6363, because of the redshift.

Zoom In Zoom Out Reset image size

The HaRs have large (>2000 km s⁻¹) recessional velocities. The recessional velocity of NGC 4388 is also large, 2524 km s⁻¹ (Lu et al. 1993), and that of the H i tail is 2000–2550 km s⁻¹ (Oosterloo & van Gorkom 2005). As the recessional velocity of the Virgo Cluster is 1079 km s⁻¹ (Ebeling et al. 1998), the peculiar velocity of NGC 4388 and its tail is >900 km s⁻¹. In the 48×36 arcmin region shown in Figure 1, the number of galaxies that have recessional velocities larger than 1500 km s⁻¹ in NASA/IPAC Extragalactic Database (NED) is seven. Two of them, AGC 226080 and VCC 956, have an H i redshift only. As the beam of the H i observation is large enough (3.3 × 3.8 arcmin; Giovanelli et al. 2007) to include the H i tail of NGC 4388, the high velocity could be measured due to an accidental overlap with the tail. The recessional velocity of VCC896 is uncertain, since it is inconsistent among Conselice et al. (2001), the Sloan Digital Sky Survey Data Release 7 (SDSS-II DR7; Abazajian et al. 2009)¹¹ and SDSS-III DR9 (Ahn et al. 2012).¹² The remaining four have optical spectroscopic redshifts that are consistent among the literature. We marked the four in Figure 1. The HaRs are very likely to be associated with the H i tail of NGC 4388 because of the large recessional velocity and their position; HaR-1 and HaR-2 overlap the H i and HaR-3 and HaR-4 lie near the contour edge (Figure 1); the nearest galaxy with a large recessional velocity is NGC 4388.

HaR-1 is a point source and was only detected in narrowband data. It is marginally detected in the near-ultraviolet (NUV) and in optical broadbands that include strong emission. Since the seeing in the NA659 filter was about 0.75 arcsec, which corresponds to 60 pc, the size of HaR-1 would be smaller than 60 pc.

HaR-2 shows a blobby and elongated appearance. Its size is 530 × 210 pc. HaR-2ab and HaR-2d were recognized separately in optical and infrared images, while they were merged in UV images due to the 6 arcsec resolution of the Galaxy Evolution Explorer (GALEX).

HaR-3 had a slightly elongated shape toward HaR-4 and showed a sign of a tail with a blob. The tail was contaminated by the light from the neighboring background galaxy and the length of the tail was uncertain. The size of HaR-3, including the tail, was >300 × 180 pc.

HaR-4 showed a possible multi-core or elongated core surrounded by a blob and extended plume (Figure 3). The size, including the plume, was 380 × 360 pc. It is interesting that HaR-3 had an Hα tail toward NGC 4388. In HaR-2ab and HaR-4, the tail/plume was on the far side of NGC 4388. The distance between HaR-3 and HaR-4 is 13 arcsec (1 kpc) and the difference in the recessional velocities of HaR-3 and 4 is 30 km s⁻¹. Even if the masses of HaR-3 and 4 are ∼10⁶ M_☉, they would not be bound gravitationally.

Internal extinction was estimated from the Balmer decrement Hβ/Hα in our FOCAS spectra. We adopted the formula by Calzetti et al. (1994),

$$\begin{equation} E(B-V)\simeq 0.935 \: \ln ((H\alpha /H\beta)/2.88). \end{equation} \tag{ 1 }$$

The estimated internal extinction is listed in Table 6.

Table 6. Hα Luminosity

Name	m	E(B − V)	log(f(Hα)	log(L(Hα)
	(NA659)		(erg s−1 cm−2))	(erg s−1))
HaR-1	23.51 ± 0.05	0.59$^{+0.13}_{-0.15}$	−16.63$^{+0.15}_{-0.13}$	35.89$^{+0.15}_{-0.13}$
HaR-2a	21.06 ± 0.04a	0.31 ± 0.07	−15.45 ± 0.07	37.07 ± 0.07
HaR-2b	...	0.23$^{+0.10}_{-0.11}$	...	...
HaR-2c	...	0.3$^{+0.2}_{-0.3}$	...	...
HaR-2d	22.45 ± 0.04b	>0.4	>−16.53	>35.99
HaR-2	...	...	...	>37.10
HaR-3	20.19 ± 0.04	0.40 ± 0.04	−15.17 ± 0.04	37.35 ± 0.04
HaR-4	19.87 ± 0.04	0.23$^{+0.07}_{-0.08}$	−14.99$^{+0.08}_{-0.07}$	37.53$^{+0.08}_{-0.07}$

Notes. ^aHaR-2a magnitude and flux includes HaR-2b and part of the 2c region. ^bHaR-2d magnitude and flux includes part of the 2c region.

Download table as:  ASCIITypeset image

In our observations, the seeing was comparable to the slit width. We therefore did not try an absolute spectrophotometric calibration from the spectra only. Instead, the spectra were used to convert the NA659 magnitude to the total flux of Hα. The extinction-corrected Hα luminosities are given in Table 6.

Kewley & Dopita (2002) presented model curves to estimate the oxygen abundance (log(O/H)+12) using several combinations of emission lines with various values of the ionization parameter (q). We can use [N ii]/[S ii], [N ii]/Hα, and [N ii]/[O iii] to estimate the metallicity. [N ii]/[O iii] was corrected for the internal extinction using Hβ/Hα and other two ratios were not corrected. q was also estimated from the curves.

From the fitting, the metallicity of the regions was estimated as log(O/H)+12 ∼ 8.6–8.7, comparable to a solar abundance (Grevesse et al. 2010). Yoshida et al. (2004) reported that the metallicity of Hα-emitting filaments near NGC 4388 was almost solar. Their ionization parameters correspond to log(U) = −2.6, −3.5, −3.2, and −3.5, for HaR-1, 2a, 3, and 4, respectively, where U is the ratio of the ionization photon density to the electron density. Since log(U) decreases with age in H ii regions, a relatively large value for HaR-1 implies that HaR-1 would be younger.

From Figure 5.8 of Osterbrock & Ferland (2006), O'Dell et al. (2013) derived the formula to estimate the electron density (n_e):

$$\begin{equation} \log (n_e) = 4.705 -1.9875 \: ({\rm [S\,{\scriptsize{II}}]6716/[S\,{\scriptsize{II}}]6731}), \end{equation} \tag{ 2 }$$

which was applicable in the regime 0.65<[S ii]6716/[S ii]6731<1.3. In this study, the error of [S ii]6717/[S ii]6731 was large and all regions except for HaR-1 were consistent with n_e ≲ 10² cm⁻³. HaR-1 may have a higher electron density, n_e ∼ 10³ cm⁻³.

We calculated model magnitudes and Hα fluxes at different ages and compared them with the observations. The details of the model are given in Appendix B. We plotted the model age versus the model mass to reproduce the observations in Figure 8. The abscissa is the logarithmic age from the burst and the ordinate is the required stellar mass of the HaR to reproduce the observed magnitude/luminosity at each age. Because the time evolution of the Hα luminosity and the UV, optical, and mid-infrared magnitudes are different, the crossing point of the tracks will indicate the stellar mass and the age.

Zoom In Zoom Out Reset image size

Hα, IR, and UV were comparable around log(T) = 6.8 and log(mass) = 4.3 in the Geneva model for HaR-2, HaR-3, and HaR-4, while i is offset by ∼0.4 dex, which corresponds to ∼1 mag and a factor of ∼2.5. Although the consistency in the Padova model was not as good as that of the Geneva model, the best-fit mass and age were comparable.

The disagreement of the estimated mass and age among the data might be explained partly by the difference in aperture sizes. Since the UV resolution was the poorest, it required the largest aperture (r = 6 arcsec), which may have included contaminants, thereby making the observed luminosity larger and shifting the lines in Figure 8 upward. However, the infrared (IR) had a smaller aperture (r = 3 arcsec) than the UV and Hα was based on NA659 photometry, which used an aperture of 2.5 × Kron radius (rK). 2.5rK was 2 ± 0.3 arcsec in the data and therefore was the smallest aperture because of the best seeing. Suprime-Cam i-band and Kitt Peak National Observatory (KPNO) mosaic ha-band fluxes were also measured in an aperture of 2.5rK radius, which was 2–3 arcsec. Nevertheless, the Hα, UV, and IR data showed good agreement, while the i-band and KPNO ha-band showed an offset in HaR-2 and HaR-3. As the i-band magnitude of Suprime-Cam was comparable to the SDSS photometry for HaR-4, which was based on the Petrosian radius, and the KPNO photometry of the ha-band showed a similar trend, the disagreement could not be caused by photometric error.

The disagreement also could come from model uncertainties and adopted assumptions, such as initial mass function (IMF), instantaneous star formation, and/or the dust extinction model. A possible model is to divide the region into several sub-regions and to assume different extinctions. For example, if some part has a large extinction, the UV is suppressed and the optical and IR fluxes could be larger relative to UV and Hα. In fact, HaR-2 consisted of sub-clumps with various extinctions. Although we can consider various star formation histories and models, such an analysis is beyond the scope of this study. We do not further tune the star formation histories and the models.

In Figure 8, HaR-1 shows the large disagreement among the data. Since HaR-1 would have a smaller mass, which was expected from its fainter i-band magnitude, it may have suffered from stochastic IMF effects (e.g., Koda et al. 2012).

As shown in Table 4, HaR-1 and 2 are 66 kpc away from NGC 4388 and HaR-3 and 4 are 35 kpc away from NGC 4388. Oosterloo & van Gorkom (2005) and Vollmer (2009) assumed that the age of the H i cloud is ∼200 Myr and estimated the projected velocity of NGC 4388 to the cloud to be 500 km s⁻¹. If we adopt this value as the velocity of the HaRs relative to NGC 4388, it should take 130 Myr (HaR-1 and 2) and 70 Myr (HaR-3 and 4) to reach the corresponding distances. Since the timescales are longer than the stellar age estimated above, the result suggests that the star formation in HaRs began after the gas left the host galaxy. This is consistent with the prediction by Yamagami & Fujita (2011).

Table 7 summarizes the literature concerning a stripped tail in nearby clusters of galaxies. Since the environment and physical condition of the tails (the ambient gas temperature and density, the relative velocity, the density of the tail, etc.) vary, we cannot compare them directly. We can at least say that HaR-1 and 2 are the most distant star-forming regions and HaR-1 may be the least massive region. Also, we can say that 10^4–5 M_☉ and ∼10⁷ yr star-forming regions in a stripped tail are not peculiar properties.

The ages of the regions in Fumagalli et al. (2011) are substantially larger than the others, probably because a different model of the star formation history was adopted. The mass estimation is not affected by the different star formation history. The detection of the star-forming regions was based on Hα and/or UV emission and therefore the sampling should be biased toward younger ages.

The detailed heliocentric velocity distributions of H i¹³ showed that the intensity of H i in the HaR-1,2 fields has a wide peak around a recessional velocity of ∼2240 km s⁻¹ (intensity-weighted mean) with a width of ∼150 km s⁻¹. HaR-1 and 2 are just on the peak. This means that the star-forming regions have only small peculiar velocities relative to the ambient H i cloud along the line of sight. Meanwhile, H i in the HaR-3,4 fields showed no significant peak. Since the star formation in the stripped tail is thought to occur in the highest-density regions (Kapferer et al. 2009), it is puzzling that the H i flux in the HaR-3,4 fields is low.

As we reported in Paper I, the tail of NGC 4388 is accompanied by relatively cool X-ray gas and the distribution of X-ray emission showed a marginal enhancement downstream of HaR-3,4. The H i gas around HaR-3 and 4 might already have evaporated, except for the densest region. Another possibility is that such dense clouds are less affected by ram pressure and left behind the H i gas (Yamagami & Fujita 2011).

The 2D spectrum and the slit profile of HaR-2 (Figures 4 and 5) shows that the Hα-strong region (HaR-2a) appears to the left of the continuum regions (HaR-2b). They resemble the fireballs found in the ram-pressure stripped tail in RB199 in the Coma cluster (Yoshida et al. 2008, 2012).

If the gas cloud (HaR-2a, 2c) is now free from the ram pressure, the gas and stars would move similarly. On the other hand, because the ram pressure affects the stars and gas differently, it would cause differential movement between the gas and the stars. The positional shift between the gas and the stars implies that ram-pressure deceleration is still working here, 66 kpc away from the host galaxy. Another possible explanation for the offset is that the gas has been consumed from the leading edge of the cloud and that the star formation is propagated to the following regions.

In HaR-4, a plume also exists on the other side of the parent galaxy. In HaR-3, however, the tail appears on the side of the parent galaxy. The direction was coincident with the direction to HaR-4. The projected distance between HaR-3 and HaR-4 was about 1.0 kpc and the difference of their recessional velocities was 30 km s⁻¹. If their masses are ∼10⁵ M_☉, the velocity difference is so large that they cannot be gravitationally bound. If HaR-3 and HaR-4 had experienced an encounter in the past, it means that such a small-scale motion existed in the tail.

Illuminated by the stars of the M86 envelope, a dark filamentary cloud was recognized around HaR-1 and HaR-2. A high contrast V-band image is shown as Figure 9. In the optical bands of Figure 2, the cloud was also recognized. The dark cloud shows an elongated and twisted shape. The length is ∼110 arcsec (7 kpc) and the width is ∼7 arcsec (600 pc) around HaR-2.

Zoom In Zoom Out Reset image size

It is uncertain whether the dark cloud is associated with the NGC 4388 tail, is an accidental overlap of a dark cloud of M86, or even if it is another dark cloud. However, as filaments of the dark cloud are only recognized along the H i tail of NGC 4388, it is likely that the dark cloud is associated with the NGC 4388 H i tail. These filaments of the dark cloud do not show any Hα emitting regions except those with HaR-1 and HaR-2. As there are only a few star-forming regions along the tail, it is unlikely that the dust was created in the tail; the origin of the dust should be NGC 4388. It implies that the dust was affected by the ram pressure. The ram-pressure stripping of dust from disk of galaxies in the Virgo Cluster has been suggested by several authors; NGC 4402 by Crowl et al. (2005), NGC 4438 by Cortese et al. (2010a), and NGC 4330 by Abramson et al. (2011). Cortese et al. (2010b) studied H i-deficient galaxies in the Virgo Cluster and found that dust is stripped in the cluster environment, as well as the gas. Yoshida et al. (2012) demonstrated the coincidence of the ram-pressure stripping of H i, Hα, and the dust in IC 4040. The elongated and twisted shape of the dark cloud downstream of HaR-1 and HaR-2 may also be explained as a result of the ram pressure that is still effective around the regions.

In Paper I, we showed evidence that M86 is more distant than the NGC 4388 tail. The total extinction of the cloud was estimated by assuming that the background M86 light was smooth. The estimated extinction at a region between HaR-1 and HaR-2 was ∼15% and ∼10% in the V-band and in the R-band, respectively, corresponding to E(B − V) ∼ 0.05. As the E(B − V) estimated from the Balmer decrement in HaR-1 is much larger (E(B − V) = 0.59$^{+0.13}_{-0.15})$, we suggest that the dust in the dark filament is locally dense around the star-forming regions.

Based on the argument of Yamagami & Fujita (2011), we discuss the star formation in the molecular clouds that later shine as the observed Hα-emitting regions. Elmegreen & Efremov (1997) derived the relation between the mass of a cloud M_c and the star formation efficiency when the cloud is under a pressure P_c. We estimate in Paper I that the ram pressure on NGC 4388 and the clouds should be P_ram ∼ 4 × 10⁻¹² dyn cm⁻¹². In Section 3.6, we found that the total mass of stars in each Hα emitting region is M_* ∼ 10⁴–10^4.5 M_☉. Assuming that P_c = P_ram, the mass of the molecular cloud before the stars were born is M_c ∼ 10^5.5 M_☉ because ∼ 0.1 (the lower part of Figure 4 in Elmegreen & Efremov 1997). For this mass and pressure, the timescale of star formation in the cloud is τ_form ∼ 10⁷ yr (the upper part of Figure 4 in Elmegreen & Efremov 1997; they referred to the timescale as the "disruption time"), which is consistent with the age of the stars in the regions (∼10^6.8 yr).

Yamagami & Fujita (2011) also discussed the disruption of the clouds by the Kelvin–Helmholtz (KH) instability via interactions with the ambient ICM. The relation among M_c, P_c, and the cloud radius R_c before star formation starts is given by

$$\begin{equation} \frac{M_c}{R_c^2}\sim 190\:{M_{\odot }\: {\rm pc}^2} \left(\frac{P_c}{1.38\times 10^{12}\rm \: erg\: s^{-1}}\right)^{1/2} \end{equation} \tag{ 3 }$$

(Elmegreen 1989). For the M_c and P_c values that we estimated, the cloud radius was R_c ∼ 43 pc. The timescale of the development of the KH instability is

$$\begin{equation} \tau _{\rm KH}=\frac{(\rho _c+\rho _{\rm ICM})R_c}{(\rho _c\rho _{\rm ICM})^{1/2}v_c}, \end{equation} \tag{ 4 }$$

where $\rho _c=3M_c/(4\pi R_c^3)$ is the density inside the cloud and v_c is the relative velocity between the cloud and the ICM (Murray et al. 1993). We estimated in Paper I that ρ_ICM ∼ 2.1 × 10⁻²⁸ g cm⁻³ and thus we obtain τ_KH ∼ 1.5 × 10⁷ yr. Since τ_KH ∼ τ_form, the cloud could marginally produce the stars before the KH instability develops. This may indicate that no magnetic fields are required to protect the cloud from the development of the KH instability (Yamagami & Fujita 2011).

From the Hα luminosity, the Hα ionizing photon emission rate of HaR-1 was 10^47.41 s⁻¹. Assuming Case B recombination, the logarithm of the total hydrogen ionizing photon emission rate Q_H was ${\rm log}(Q_H) \sim 47.8^{+0.15}_{-0.13}$. According to Sternberg et al. (2003), a single B0V dwarf is enough to produce that photon rate; other stars such as B giants and O dwarfs would make too much Hα.

However, a single B0V star is insufficient to reproduce the observed NUV magnitude. The absolute V magnitude of a B0V star is −4 (e.g., Wegner 2006). The NUV-V color of B0V was calculated with empirical spectral energy distributions (SEDs) by Pickles (1998) as −1.0 mag. The absolute NUV magnitude is therefore ∼−5. Meanwhile, the absolute magnitudes of HaR-1 converted from observed NUV magnitudes and the internal extinction correction was −8.8±0.7 mag. The observed UV magnitudes were >3 mag brighter.

The UV–optical color was also inconsistent with the following models. We calculated the NUV − i color of various SEDs and found that the bluest stars are O9V–B0V (NUV−i ∼ −1.7). The observed lower limit of the NUV−i color of HaR-1 is NUV − i < −1.6 and if we correct for the internal extinction of E_gas(B − V) = 0.59, the color is NUV − i < −2.9. The model SEDs, the internal extinction estimated from the Balmer decrement, the extinction law, and the factor of 0.44 are therefore inconsistent with the observed color.

The Suprime-Cam data used in the study are summarized in Table 8. The data were retrieved from SMOKA.¹⁷

The data were reduced the same way as in Yoshida et al. (2002). As the i-band data were obtained with FDCCDs, crosstalk corrections were applied (Yagi 2012). The World Coordinate System was calibrated using wcstools 3.8.3 (Mink 2002) and the Guide Star Catalog 2.3 (Lasker et al. 2008). Photometric zero points were calibrated using the method given in Yagi et al. (2013). We adopted the stars in SDSS-III DR 9 as standard objects and converted the SDSS magnitudes to the Suprime-Cam system. The details of the construction of the magnitude conversion were described in Yagi et al. (2013). The color conversion coefficients from SDSS to Suprime-Cam magnitudes are given in Table 9. Stars of 19 < r < 21 magnitude in SDSS3 DR9 were used for our calibration. The K-correct (Blanton & Roweis 2007) v4 offset for SDSS¹⁸ is applied; m_AB − m_SDSS = 0.012, 0.010, and 0.028 for g, r, and i, respectively.

In Figures 2 and 3, Hα-only images (NB659-R) are obtained by subtracting the scaled R-band image from the NA659-band image. As the seeing was worse in the R-band, the NA659 image was convolved with a Gaussian (FWHM = 0.95 arcsec) to match the size of the point spread function (PSF).

We used SExtractor (Bertin & Arnouts 1996) for the photometry of the Suprime-Cam data. The Galactic extinction was corrected for using the values in NED based on Schlegel et al. (1998) and Schlafly & Finkbeiner (2011): (A_V, A_R, A_i, A_NA503, A_NA659) = (0.08, 0.06, 0.05, 0.09, 0.06) and (0.08, 0.06, 0.05, 0.10, 0.07) for the HaR-1,2 field and the HaR-3,4 field, respectively. Their magnitudes (MAG_AUTO) after the Galactic extinction correction are given in Tables 2 and 3.

The NOAO science archive¹⁹ provides calibrated data around the region that were taken on 2007 April 22 and 2007 April 23 with NOAO MOSAIC 1 on the KPNO 4 m Mayall telescope. The data were already used in Kenney et al. (2008). We retrieved the data in rest-Hα (ha; center = 6566 Å, FWHM = 80 Å at F = 3.1), Hα around the Virgo redshift (ha4; center = 6610.1 Å, FWHM = 80 Å at F = 3.1), and the R-band (center = 6622 Å, FWHM = 1490 Å at F = 3.1). The filter characteristics are given on the KPNO Web site,²⁰ and the center and FWHM were calculated from the transmission data and a 15 Å shift of the R-band at F = 3.1.

SExtractor (Bertin & Arnouts 1996) was used for photometry. The photometric zero points were given in the FITS header. We adopted the Galactic extinction of 0.06 for HaR-1 and HaR-2 and 0.07 for HaR-3 and HaR-4. The AB-Vega offset of 0.25 mag was also applied. The results are listed in Table 3.

In the GALEX archive,²¹ several coadded tiles covered the targets. We retrieved the coadded FITS images listed in Table 10 and measured the flux in a circular aperture in the FUV (1344–1786 Å) and NUV (1771–2831 Å) bands (Morrissey et al. 2007). As the FWHM of the PSF of GALEX was large, we adopted an aperture size 6 arcsec in radius. The files presented as *_skybg.fits of each tile was used for the background correction. The background-subtracted aperture fluxes in the tiles were averaged weighted by exposure time and converted to AB magnitudes using the zero point given in Morrissey et al. (2007). We adopted the Galactic extinction correction in the GALEX bands given by Wyder et al. (2007): A_FUV = 8.24E(B − V) and A_NUV = 8.24E(B − V) − 0.67E(B − V)². The results were A_FUV, A_NUV = (0.206,0.206), (0.230,0.231) for the HaR-1,2 fields and the HaR-3,4 fields, respectively. The magnitude errors were estimated following the GALEX Web site.²² The results are given in Table 2. In the FUV data, HaR-1 shows no recognizable shape. We therefore assumed the measured aperture flux as the upper limit.

The target fields were also observed by the XMM-Newton Optical Monitor (XMM-OM) and UVW1 (center = 2905 Å, width = 620 Å; Kuntz et al. 2008) magnitudes of HaR-2,3, and 4 are available in the catalog in the archive. The data were obtained from MAST.²³ The catalog magnitudes are listed in Table 11. Although the error on the magnitude was larger than that of GALEX, the spatial resolution of XMM-OM was higher: the FWHM of the PSF was ∼5 arcsec and ∼2.3 arcsec for the GALEX NUV band and the XMM-OM UVW1 band, respectively. We checked that there is no apparent contamination from nearby objects around the targets in the UVW1 band of XMM-OM.

HaR-4 was detected in SDSS3 DR9 as SDSS J122613.13+ 124300.4. The SDSS magnitudes are listed in Table 12. We corrected for Galactic extinction and the offset of SDSS magnitudes from AB magnitudes using K-correct, version 4 (Blanton & Roweis 2007); m_AB − m_SDSS = −0.036, 0.012, 0.010, 0.028, and 0.040 for the u, g, r, i, and z-bands, respectively. The i-band magnitude was comparable to our measurement in Suprime-Cam in Table 2. The other three HaRs were not detected in SDSS.

We found that the model magnitude of HaR-4 was ∼0.7 mag brighter in SDSS DR7 (Abazajian et al. 2009) than that in DR9. The reason could be the difference in the adopted model profile of the object: the Petrosian radius was 7.359 arcsec (DR7) versus 2.475 arcsec (DR9). As seen in the bottom of Figure 3, whose image size is 20 arcsec on a side, the Petrosian radius of 7.3 arcsec was too large for HaR-4 and we considered the magnitude in DR9 to be appropriate.

The ∼0.7 mag difference was specific to HaR-4. Other brighter objects had comparable magnitudes. Our photometric calibration of the Suprime-Cam data based on the DR9 data was not affected even if we used the DR7 data.

We retrieved calibrated data from the Infrared Array Camera (IRAC) on the Spitzer Space Telescope in the 3.6 μm (ch1) and 4.5 μm (ch2) channels from IRSA.²⁴ Five images taken in 2010 February and March covered our target fields. Their unique observation identification numbers (AORKEY) and observation dates are listed in Table 13. The exposure times were 93.6 and 96.8 s for ch1 and ch2, respectively.

We used the flux zero point given in the calibrated FITS headers (1 DN pixel⁻¹ = 10⁶ Jy sr⁻¹). After background subtraction, aperture photometry with a 3 arcsec radius was performed for each target in each image. Then, the median flux was calculated for each target.

The error of each image was estimated from the median of the absolute deviation of random aperture photometry assuming a 3 arcsec radius. AB magnitudes were calculated as

$$\begin{equation} {\rm AB}=23.90\hbox{--}2.5\log (S(\mu {\rm Jy})). \end{equation} \tag{ A1 }$$

The results are listed in Table 2.

HaR-1 was not detected in the IRAC images and the aperture flux was negative in some of the images. The upper limit of the aperture flux in a 3 arcsec radius in ch1 (0.7 μJy) was therefore used. HaR-3 was heavily blended with the neighbor galaxy. We tried to measure the flux in a semi-circular aperture to mask the side of the neighbor galaxy and double the flux. The value should still suffer from the envelope of the neighbor galaxy and we regard it as the upper limit of the flux.