SPECTACULAR X-RAY TAILS, INTRACLUSTER STAR FORMATION, AND ULXs IN A3627

The observation of ESO 137-001 was performed with the Advanced CCD Imaging Spectrometer (ACIS) on 2008 June 13–15 (obsID: 9518). Standard Chandra data analysis was performed which includes the corrections for the slow gain change and charge transfer inefficiency (for both the FI and BI chips). We examined the light curve from the part of the S3 chip with weaker ICM emission. No flares of particle background were found. There are also no flares in the light curve of the I2+I3 chips. The effective exposures are 138.2 ks for the BI chips and 140.0 ks for the FI chips. This new Chandra exposure is 18 times deeper than the old one. About 4100 net counts from the thermal gas were collected in the 0.5–3 keV band. We corrected for the ACIS low-energy quantum efficiency (QE) degradation due to the contamination on ACIS's optical blocking filter, which increases with time and varies with position. The dead area effect on the FI chips, caused by cosmic rays, has also been corrected. We used CIAO3.4 for the data analysis. The calibration files used correspond to Chandra calibration database (CALDB) 3.5.3 from the Chandra X-ray Center. We used the new HRMA on-axis Effective Area file, "hrmaD1996-12-20axeffaN0008.fits," which improves the accuracy of the temperature measurement at kT > 4 keV. The remaining uncertainty in the source flux (up to ∼10%) would not much affect our conclusions in this work. In the spectral analysis, lower energy cutoffs of 0.4 keV (for the BI data) and 0.5 keV (for the FI data) are used to minimize the effects of calibration uncertainties at low energy. The solar photospheric abundance table by Anders & Grevesse (1989) is used in the spectral fits. Uncertainties quoted in this paper are 1σ. We used Cash statistics in this work. We adopted an absorption column density of 1.73 × 10²¹ cm⁻² from the Leiden/Argentine/Bonn H i survey (Kalberla et al. 2005). If we choose to fit the absorption column from the Chandra spectra (with the PHABS model in XSPEC), the best fits are always consistent with the above value. This absorption column is lower than the previous value from Dickey & Lockman (1990), 2.0 × 10²¹ cm⁻², which had been used in previous work (e.g., Böhringer et al. 1996; S06).

We obtained optical spectra of the emission-line objects identified by S07 with GMOS on the Gemini South (program ID: 2008A-Q50). Because the field is crowded with interesting sources, three masks had to be used with some sources covered in more than one mask. Observations with the first mask were done on 2008 May 3 (airmass: 1.18–1.22, seeing: ∼09). Two exposures (20 minutes each) were taken, with a central wavelength difference of 60 Å in order to cover the inter-chip gaps. Observations with the second mask and the third mask were taken on 2008 June 26 and June 27, respectively (airmass: 1.17, seeing: ∼2'' for June 26 and airmass: 1.18, seeing: ∼13 for June 27). Two exposures (10 minutes each) were taken on each night. All three nights were photometric. The R400 grating + GG455 filter were used, which gives a free wavelength range of ∼4500–8000 Å. We used 1'' slits for all sources. As queued observations, we used the calibration data of LTT4364 taken on the photometric night of 2008 May 27 for the relative flux calibration.

The GMOS data were reduced with the Gemini IRAF Package (ver. 1.9.1). We generally followed the standard procedures. Besides the accompanying CuAr arc data, we also used the night sky lines for the wavelength calibration. The latter procedure is the key to derive consistent velocities for the same source from different exposures. In this paper, we present only the velocity information and briefly discuss the nature of these sources (H ii regions). The detailed properties of these H ii regions will be discussed in a future paper.

The observed soft X-ray emission of the tails most likely comes from the interfaces between the hot ICM and the cold stripped ISM. Although there are no data for the cold atomic and molecular gas (note the Spitzer data on the polycyclic aromatic hydrocarbon (PAH) emission and the warm molecular H₂; Sivanandam et al. 2010), the X-ray surface brightness of the tails reveals a great deal about the stripped ISM. The surface brightness of the tails is quantitatively examined. We first derived the 0.6–2 keV surface brightness profiles across some parts of the tails and the leading edge. As shown in Figure 3, both tails are narrow but highly significant features above the local background. Even the "protrusion" is a 4.1σ feature. The leading edge (or the contact discontinuity) is very sharp, corresponding to the Hα edge (Figure 4). The gap between two tails has a similar surface brightness as the local background (Figures 3(b) and (c)). For the front part of the gap, the surface brightness is (6.26 ± 0.30) × 10⁻⁵ counts s⁻¹ kpc⁻², while the local background is (5.88 ± 0.27) × 10⁻⁵ counts s⁻¹ kpc⁻². The low surface brightness of the gap has implications for the three-dimensional structure of the stripped ISM tail. One possibility for the observed double tails is the projection effect of one broad cold ISM tail. According to this interpretation, the three-dimensional structure of the soft X-ray emission might be a cylindrical shell that produces double tails in the projected surface brightness. As shown in Figure 3, such a configuration fails to simultaneously account for both the low surface brightness of the gap and the observed widths of two tails. The observed emission level between tails is similar to that outside tails. Basically, the width of the cylindrical shell required to make the central emission dim also makes each peak unacceptably narrow in relation to the observed profiles. Thus, the two X-ray tails are most likely detached. This is also supported by the Spitzer data (Sivanandam et al. 2010). The 8 μm PAH emission only appears in the first 20 kpc of the main tail. The IRS data only cover a small field downstream but no enhanced warm H₂ emission is present at the front part of the gap. Eventually, we need H i and CO data to better reveal the distribution of the cold ISM downstream of ESO 137-001.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In principle, the surface brightness profiles along the tails reflect the stripping history. As shown in Figure 5, both tails have wiggles and substructures along the (time) axis. However, the unknown conversion from the X-ray emission to the mass of the stripped ISM (filling factor, detail of the emission spectra, etc.) makes this difficult with only the X-ray data. A more complete picture can be obtained with the H i and CO data. From Figure 5, the total linear length of the whole tail system is at least 42 or 80 kpc. We also measured the widths of the main tail in four segments (before sharp bends) and the widths of the secondary tail in two segments. The model is composed of a Gaussian and a flat local background. As shown in Figure 6, the widths of both tails do not increase with the distance from the galaxy. This conflicts with simulations (e.g., RB08; Kapferer et al. 2009), as further discussed in Section 7.2. If ESO 137-001 is moving supersonically, a bow shock may form. The local sound speed is 1263 (kT/6 keV)^1/2 km s⁻¹. For a Mach number of 1.1–3, the predicted position of the bow shock is 20''–2'' from the leading edge (or the contact discontinuity). No such shock front is detected in the Chandra exposure. The enhanced surface brightness would be much weaker in projection on the sky, especially if ESO 137-001 is farther from the cluster center than the projected position.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

With this deep Chandra exposure, for the first time spatially resolved spectroscopy can be done for X-ray stripping tails of a cluster late-type galaxy. We measured gas temperatures in five regions of the main tail and two regions of the secondary tail (Figure 7). The background is from the local source-free regions. A single APEC model with Galactic absorption was used. We emphasize that the measured temperatures are only spectroscopic temperatures as the intrinsically multi-phase gas is fitted with a single-kT model. In fact, a hard X-ray excess generally exists in the spectral fits, which implies the existence of hotter components. However, with limited statistics, we elected to keep the spectral model simple. As shown in Figure 7, the variations of spectroscopic temperatures along both tails are small. In fact, both tails have consistent temperatures.

Zoom In Zoom Out Reset image size

Since temperatures derived from individual regions are similar, the total spectrum of two tails (still excluding the "head" region) was also examined with several models (Figure 8 and Table 1). This observation collects 3444 net counts from the thermal gas of the two tails in the 0.5–3 keV band. While the fit with a single-kT model is acceptable, the addition of a second component (either another thermal component or a power-law component) produces better fits. We also tried the XSPEC model CEMEKL, which allows a power-law distribution of the emission measure for multi-kT gas. We consider the spectral results shown in Table 1 as evidence for multi-phase gas, especially as the single-kT model produces a very low abundance. Adding another or more thermal components can increase the abundance to the solar value. Nevertheless, the double APEC or the CEMEKL models are no more than a phenomenological way to describe the data. It is still unclear what emergent spectrum mixing will produce. There are more complications, like the non-equilibrium ionization effect and the intrinsic absorption. Generally, ionization equilibrium requires the collisional ionization parameter n_et > 10¹² cm⁻³ s. For n_e = 10⁻²–10⁻³ cm⁻³, t > 3–32 Myr, which is comparable to the age of the X-ray tails.

Zoom In Zoom Out Reset image size

The X-ray gas mass of the tails can be estimated. The main uncertainties are the geometry (or the filling factor) of the X-ray emitting gas and the abundance of the gas. The filling factor of the soft X-ray emitting gas depends on the mass and configuration of the stripped cold ISM gas in the tails. The soft X-rays may originate from the interfaces between the ambient hot ICM and the stripped cold ISM. A fragmented cold ISM tail has larger surface area, so the filling factor of the soft X-ray emitting gas would be larger. Both uncertainties can be constrained by the assumed pressure equilibrium with the ambient ICM. In the following, we present the estimates from the one-kT fit, the two-kT fit, and the CEMEKL fit. We assume both the main tail and the secondary tail are cylinders with a length of 80 kpc and radii of 3.9 kpc and 3.4 kpc, respectively. The "head" and "protrusion" regions are not included. With the result from the one-kT fit (the first row of Table 1), the total X-ray gas mass is 2.0 × 10⁹ f^1/2 M_☉, where f is the filling factor of the soft X-ray emitting gas. If the estimate is done on the same individual regions as the spectral analysis (Figure 7), the total gas mass only decreases by 3%. The electron density ranges from 0.0080 f^−1/2 cm⁻³ to 0.017 f^−1/2 cm⁻³ along the tails. However, the abundance of the X-ray gas is too low for the one-kT fit, which results in gas densities that are too high. The ambient ICM electron density and temperature in the projected position of ESO 137-001's tails is ∼1.3 × 10⁻³ cm⁻³ and 6.3 keV, respectively. This puts an upper limit of 0.010 cm⁻³ on the electron density of the tail gas (for a temperature of 0.8 keV) if pressure equilibrium is assumed. Thus, the one-kT model is not sufficient.

Do the two-kT fit and the CEMEKL fit produce more reasonable results? We first used the result in the fourth row of Table 1, where the abundance of the lower temperature component is fixed at the solar value, and the abundance of the higher temperature component is fixed at the value of the surrounding ICM. Assuming pressure equilibrium between these two components, it is found that the hotter component occupies 97% of the volume of the soft X-ray emitting gas. The electron density of the cool and hot components is 0.018 f^−1/2 cm⁻³ and 0.0071 f^−1/2 cm⁻³, respectively. The total X-ray gas mass is 1.4 × 10⁹ f^1/2 M_☉. However, the gas pressure in the tail is ∼50% higher than the maximum ICM pressure of the surroundings. This problem cannot be solved by increasing the abundance of the hotter component in the spectral fit, which will further increase the ISM pressure. We then tried the CEMEKL fit (the eighth row of Table 1). However, the X-ray tail is at least 3.2 times overpressured to the surrounding ICM, under an isobaric condition. The total X-ray gas mass is 1.1 × 10⁹ M_☉. Therefore, either these models are not adequate, or there is dynamic pressure (e.g., turbulence) besides the thermal pressure. Without better understanding of the tail spectrum and the contribution of the cold ISM in the tails, it is difficult to much improve the above estimates. We summarize ESO 137-001's properties from the current data in Table 2, along with those of ESO 137-002 (Section 6).

S07 identified 29 strong candidate H ii regions. Six more sources were also identified as possible H ii regions. The Gemini GMOS observations were able to cover 33 of them (solid blue squares in Figure 9). Only two sources (dashed blue squares in Figure 9) were not observed. Both of them are strong candidates but are faint and close to ESO 137-001. The GMOS spectra unambiguously confirm the association of these 33 sources (27 strong candidates + 6 weaker candidates in S07) with ESO 137-001 as their velocities are all within 200 km s⁻¹ of ESO 137-001's. The heliocentric velocity map of these 33 H ii regions is shown in Figure 10. In Figure 10, we also plot their velocities versus their distances to the major axis of the galaxy or the middle line of the bow-like front of H ii regions. We used a system velocity of 4660 km s⁻¹ for ESO 137-001, which is the weighted average from the velocities derived by Woudt et al. (2004) and S07. The velocities of these H ii regions span a small range (⩽300 km s⁻¹) with a dispersion of 54⁺¹²₋₁₀ km s⁻¹. In fact, this small-velocity range is comparable to the radial rotation velocity of the residual Hα nebula at the center of ESO 137-001, ∼80 km s⁻¹ from our SOAR GOODMAN spectra that will be presented in a future paper. After ISM is removed from the galactic disk, it is subject to turbulence in the wake. Combined with the residual rotational motion, the range of the radial velocities of stripped ISM clouds should become substantially larger than the range of the rotation velocity (e.g., RB08). However, that is not what we observed with the velocities of H ii regions as tracers. In fact, as shown in Figure 10, the imprint of the rotation velocity pattern is still present in the velocity map of these H ii regions.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Some example spectra of these H ii regions are shown in Figure 11, including several sources that are most distant from ESO 137-001. The low [O i] λ6300/Hα and [N ii] λ6584/Hα ratios are typical of giant H ii regions. The detailed properties of these H ii regions (e.g., line equivalent width, line ratios, star formation rate (SFR), age, metallicity, and total mass) will be presented in a future paper with the results from our Hubble Space Telescope (HST) data.

Zoom In Zoom Out Reset image size

The deep Chandra exposure also reveals many point sources, as shown in Figure 9. We focus on X-ray point sources in ESO 137-001 or around its X-ray tails (within the two elliptical regions in Figure 9). Nineteen point sources are detected. Their positions and fluxes are listed in Table 3. The 0.5–2 keV limiting flux is ∼3.5 × 10⁻¹⁶ erg s⁻¹ in the region of interest (with eight counts in the 0.5–2 keV band). The expected X-ray source density above this limiting flux is ∼1500 per deg² (Kim et al. 2007). For the two elliptical regions (9.66 arcmin²), the expected number of background sources is four, while 19 point sources are detected. The contribution of Galactic sources should not be neglected as ESO 137-001 is located at a low Galactic latitude (−7°). Indeed, six Chandra point sources are positionally coincident with bright star-like objects. However, an excess of X-ray point sources still exists, especially in the immediate downstream region of ESO 137-001 where most H ii regions are. We can divide these point sources into three groups, sources 1–3 that are upstream of the galaxy, sources 4–12 that are around the disk plane or downstream of the galaxy (but still within the 15 kpc radius halo in projection), and sources 13–19 that are unbound and near the X-ray tails. The latter two groups are of most interest here. The region immediately downstream of the galaxy (but still within the galactic halo) has nine X-ray sources, while only 0.5 are predicted from the log N–log S relation (∼1.1 arcmin²). The unbound tail region has seven X-ray sources, while 3.3 sources are expected from the log N–log S relation (∼7.9 arcmin²). Three sources are likely of stellar origin (14, 16, and 17), while source 15 is close to a confirmed H ii region. If we take three stars in the unbound tail region as the number density of stars, only 0.4 stars are expected in the immediate downstream region. In fact, only source 7 coincides with a bright star. Thus, the excess of X-ray point sources in the immediate downstream region is significant.

Table 3. The Properties of Chandra Point Sources Around ESO 137-001 and its Tails

Source a	R.A.b	Decl.b	0.5–7 keV Countsc	fd0.3–10 keV	Γ	Note
	(J2000)	(J2000)		(10−15 erg s−1 cm−2)
1	16:13:30.43	−60:45:18.2	17.3 ± 5.1	1.6 ± 0.5	(1.7)	A bright stellar object
2	16:13:29.76	−60:46:30.0	47.3 ± 7.6	12.4 ± 2.0	0.02 ± 0.31	A faint optical source
3	16:13:27.66	−60:46:22.7	35.4 ± 6.6	3.2 ± 0.6	(1.7)	A bright stellar object
4	16:13:27.75	−60:45:34.4	17.5 ± 5.0	1.6 ± 0.5	(1.7)	No optical counterpart
5	16:13:27.44	−60:45:54.4	41.4 ± 7.3	3.7 ± 0.7	(1.7)	No optical counterpart
6	16:13:26.03	−60:45:38.3	13.9 ± 4.6	1.3 ± 0.4	(1.7)	No optical counterpart
7	16:13:25.80	−60:45:45.3	64.6 ± 9.1	6.6 ± 1.0	3.45 ± 0.44	A bright stellar object
8*	16:13:25.76	−60:45:10.3	33.3 ± 6.7	3.0 ± 0.6	(1.7)	Between two H ii regions, 13 from them
9	16:13:24.43	−60:45:45.2	115.1 ± 11.6	9.8 ± 1.0	1.87 ± 0.17	40 from an H ii region
10*	16:13:24.30	−60:46:06.6	42.0 ± 7.2	3.5 ± 0.6	2.17 ± 0.33	20'' from an H ii region
11*	16:13:24.29	−60:45:37.2	116.5 ± 11.7	11.4 ± 1.2	1.54 ± 0.19	08 from an H ii region
12	16:13:23.57	−60:45:29.8	414.3 ± 21.0	42.6 ± 2.2	1.38 ± 0.09	26 from an H ii region
13	16:13:18.28	−60:43:55.8	32.6 ± 6.6	3.0 ± 0.6	(1.7)	No optical counterpart
14	16:13:16.94	−60:44:51.5	23.0 ± 5.7	2.1 ± 0.5	(1.7)	A bright stellar object
15*	16:13:16.76	−60:44:18.1	16.8 ± 4.8	1.6 ± 0.5	(1.7)	10 from an H ii region
16	16:13:10.74	−60:44:37.6	16.5 ± 4.9	1.5 ± 0.5	(1.7)	A bright stellar object
17	16:13:10.19	−60:45:04.0	32.7 ± 6.5	3.0 ± 0.6	(1.7)	A bright stellar object
18	16:13:07.12	−60:44:54.3	17.0 ± 4.9	1.6 ± 0.5	(1.7)	No optical counterpart
19	16:13:06.34	−60:43:38.0	48.4 ± 7.9	4.9 ± 0.8	1.38 ± 0.32	A faint optical source

Notes. ^aSources with * are considered strong candidates of intracluster ULXs. Sources 9 and 12 are also candidates. ^bThe position uncertainty is ∼04 from the examination of eight stars in the field. ^cCounts and errors are from "WAVDETECT" of CIAO. ^dWe always assumed a power law and the Galactic absorption. Spectra of bright sources were analyzed individually, while a photon index of 1.7 was simply assumed for fainter sources. The Galactic stars would have smaller absorption but they are not the focus of this work. The four promising candidates for intracluster ULXs have rest-frame 0.3–10 keV luminosities of (0.92–6.6) × 10³⁹ erg s⁻¹, if in A3627. The other candidates (at least 9 and 12) have rest-frame 0.3–10 keV luminosities of up to 2.5 × 10⁴⁰ erg s⁻¹, if in A3627.

Download table as:  ASCIITypeset image

We examined the optical counterparts of these X-ray point sources and list the results in Table 3. The SOAR images (B, I, and Hα bands) that are presented in S07 are used. With seven sources (across the S3 chip) detected in both the X-rays and the optical images, the astrometry can be determined to ∼03. There are four Chandra sources (8, 10, 11, and 15) that are within 2'' (or 0.65 kpc) of the confirmed H ii regions and have no other optical counterparts in the SOAR images. We consider them strong candidates for intracluster X-ray binaries, while a 0.5 kpc offset can be produced by a 100 km s⁻¹ kick velocity in 5 Myr. Sources 9 and 12 are 0.85–1.3 kpc from confirmed H ii regions, and do not fall on any optical sources. They are also candidates for intracluster X-ray binaries. If associated with ESO 137-001, their 0.3–10 keV luminosities are 0.92–25 × 10³⁹ erg s⁻¹ (Table 3), which make them intracluster ULXs. The X-ray spectra of the bright point sources were also examined (Table 3). It is true that the chance match between H ii regions and X-ray point sources is high in the immediate downstream region. There is a ∼52% chance that an X-ray source will be within 2'' of an H ii region there. However, the excess of X-ray sources is also significant there.

As ULXs are often found around young massive star clusters, it may not be a surprise to have intracluster ULXs around intracluster massive star clusters. We can further examine the relation between the X-ray luminosities of these putative ULXs and the total SFR of the H ii regions. The total 2–10 keV luminosity of sources 8, 10, 11, and 15 is 6.5 × 10³⁹ erg s⁻¹. If sources 9 and 12 are added, the total luminosity is 2.7 × 10⁴⁰ erg s⁻¹. From S07, the total SFR of these 33 H ii regions is ∼0.7 M_☉ yr⁻¹, assuming an intrinsic extinction of 1 mag for the Hα emission. This SFR and the lower value of the X-ray luminosity fall well on the relation derived by Grimm et al. (2003). Even for the higher value of the X-ray luminosity (when sources 9 and 12 are included), agreement can be achieved if the intrinsic extinction is higher in these H ii regions and the SFR in the past was higher.

The details of the Chandra observation of ESO 137-006 (obsID 8178) and of the data analysis were presented in Section 5 of Sun (2009). The effective exposure was 57.3 ks for the I2 chip. A significant X-ray tail behind ESO 137-002 is revealed, which in fact can also be seen in the XMM-Newton data (Figure 2).

Being aware of its X-ray tail, we observed ESO 137-002 (z = 0.0191) with the 4.1 m SOAR Telescope on Cerro Pachon on 2008 April 18 (UT). The observations were made with the SOAR Optical Imager (SOI), which covers a 52 × 52 field of view. The night was photometric and seeing was very good, 045–065. The airmass was 1.16–1.20. We took three 800 s exposures with the CTIO 6693/76 filter (Hα band) and two 600 s exposures with the CTIO 6606/75 filter (Hα_off band).⁶ Both narrowband filters are 2'' × 2'', so the non-vignetted field is only ∼36 × 35. Dithers of 15''–30'' were made between exposures. Each image was reduced using the standard procedures with the IRAF MSCRED package. The pixels were binned 2 × 2, for a scale of 0154 per pixel. Dome flats were used. The US Naval Observatory (USNO) A2.0 catalog was used for the World Coordinate System (WCS) alignment, which was done with WCSTools. The spectrophotometric standards were LTT 4364 and EG 274 (Hamuy et al. 1994).

ESO 137-002 is a disk galaxy with a dust lane across its bulge (Figure 13). It is a more massive galaxy than ESO 137-001 and has a larger radial velocity component than ESO 137-001 (Table 2). ESO 137-002 is 120 kpc (or 615) from ESO 137-001 in the plane of sky. Assuming the same ratios of the halo mass to the stellar mass, the tidal radius [r_t ∼ (m/3M)^1/3D] is 0.32–0.41 D, where D is the distance between the two galaxies. This value is much larger than the tidal truncation radius of ESO 137-001 (∼15 kpc, truncated by the cluster potential) estimated by S07. Even if the two galaxies are only separated by their projected distance, ESO 137-001's gravity is over six times ESO 137-002's within 15 kpc radius of ESO 137-001's nucleus. Given the large difference of their radial velocities (∼1080 km s⁻¹), their actual spatial separation should be larger than the projected distance, especially at the time when stripping began. Thus, the tidal interaction between two galaxies should be small. There are indeed no tidal features upstream of ESO 137-001 and around ESO 137-002. Their X-ray tails are also pointing away from each other.

ESO 137-002's X-ray tail can be traced to at least 40 kpc from the nucleus. Since it is 106 off axis and the galaxy has a large radial velocity component, the actual length of the X-ray tail may be larger. The 3–8 keV image clearly reveals a bright nuclear source. We examined ESO 137-002's X-ray spectral properties in two regions, one within 174 of the nucleus (note the 90% encircled energy radius of the point-spread function (PSF) at 1.49 keV is 14'' there) and the other one excluding the nuclear region. The first region has 460 ± 30 net counts in the 0.5–10 keV band, while the second region (the tail region) has 253 ± 29 net counts in the 0.5–2 keV band. The spectra are shown in Figure 14 and the spectral fits are shown in Table 4. ESO 137-002 hosts a Seyfert2-like nucleus. The limited statistics on the soft X-rays, combined with the poor angular resolution in the source position, prevent any definite conclusion on the gas properties. A single-kT fit implies a surprisingly hot tail with kT∼ 2 keV. The gas temperature can be reduced to ∼1.1 keV if a power-law component is included. However, the power-law component has to be very strong (Table 4), given the small area of the tail region (0.81 arcmin², compared with the case of ESO 137-001). Like ESO 137-001's tail, the abundance of ESO 137-002's gas cannot be reliably constrained. The X-ray gas mass of the tail can also be estimated. We assume the tail as a cylinder with a length of 40 kpc and a radius of 2.5 kpc. Using the result from the one-kT fit (the first row of Table 4), the total gas mass is 4.2 × 10⁸ f^1/2 M_☉, where f is the filling factor of the soft X-ray emitting gas. This may be an overestimate as the gas abundance is too low. In fact, the average electron density is 0.0188 f^−1/2 cm⁻³. The surrounding ICM has a temperature of ∼6.3 keV, similar to that around ESO 137-001. The ICM electron density at the projected position is ∼1.9 × 10⁻³ cm⁻³ (Böhringer et al. 1996). Thus, the X-ray tail would be ∼3 times overpressured. With the addition of a power-law component and a fixed abundance at 1 solar (the third fit in Table 4), the ISM electron density and the total gas mass reduce to 0.0055 f^−1/2 cm⁻³ and 1.2 × 10⁸ f^1/2 M_☉, respectively. The last spectral fit may largely underestimate the contribution from the gas (see the difference of luminosities in Table 4). Thus, we take an average mass estimate of ∼2 × 10⁸ f^1/2 M_☉.

Zoom In Zoom Out Reset image size

The net Hα image of ESO 137-002 also shows an Hα tail at the same position as the X-ray tail, extending to at least 20 kpc from the nucleus (Figure 13). A sharp Hα edge is located only 1.1 kpc from the nucleus. Hα absorption is found upstream of the edge in the disk plane. We also searched for emission-line objects in the field from the narrowband data, hoping to find similar intracluster H ii regions as those downstream of ESO 137-001. However, no such emission-line regions have been found in the sources detected in the Hα band. The upper limit on the Hα luminosity (no intrinsic extinction) is 10³⁸ erg s⁻¹, and the upper limit on the equivalent width of the Hα line is 60 Å. Assuming a metallicity of 0.4 solar and applying the Starburst99 model (Leitherer et al. 1999), this upper limit gives a lower limit of 7 Myr for the age of any faded intracluster H ii regions. This may imply that ESO 137-002 is in a more advanced stage of stripping and evolution than ESO 137-001, which is consistent with the likely hotter X-ray tail of ESO 137-002 and ESO 137-002's redder optical color. ESO 137-002 is a more massive galaxy than ESO 137-001 and may be undergoing edge-on stripping, which indeed takes longer than the face-on stripping (e.g., RB08).

While the paucity of luminous X-ray tails of late-type galaxies in nearby clusters has been known (e.g., Sun et al. 2007b), the discovery of ESO 137-001's double X-ray tails is a surprise. M86 also has double X-ray tails and Randall et al. (2008) interpreted it as the consequence of the aspherical potential of M86 and M86's orbit. M86 is an early-type galaxy so its potential structure is different from ESO 137-001's and its tail is only composed of the stripped hot ISM. M86's tails are in fact connected to the "plume" on the north of the galaxy and are located at a larger distance from the galaxy, instead of coming from the galaxy directly like ESO 137-001's. Moreover, a spectacular Hα complex connecting M86 and the nearby disturbed spiral NGC 4438 was recently discovered by Kenney et al. (2008). Some Hα emission also extends to the position of M86's tail. Thus, tidal interaction may also be important for M86's X-ray tail. There also appears to be a split feature that resembles a double tail in NGC 4438's Chandra image (Machacek et al. 2004; Randall et al. 2008). However, one part of the split feature is almost along the distorted disk plane and tidal interaction should be important in this case (Machacek et al. 2004; Kenney et al. 2008). NGC 4438 is ∼3 times more luminous than ESO 137-001 in the K_s band. Its truncated halo should be larger than ESO 137-001's (∼15 kpc in radius, S07), also because the Virgo cluster is less massive than A3627. The features in NGC 4438 only extend to ∼11 kpc from the nucleus, clearly within its halo. The other late-type galaxy in the proximity, NGC 4388, has many Hα filaments and a spectacular H i tail (Yoshida et al. 2002; Oosterloo & van Gorkom 2005). A faint X-ray extension can be traced to ∼14 kpc from the nucleus in the direction of the H i tail, from the Chandra and XMM-Newton data. There is also an X-ray extension to east in the disk plane, which creates the appearance of double tails (Randall et al. 2008). NGC 4388 is ∼1.5 times more luminous than ESO 137-001 in the K_s band so similar to NGC 4438, its X-ray tail is still within its halo. Thus, we do not consider the one-sided X-ray features in NGC 4438 and NGC 4388 to be X-ray double tails based upon the current data. They are clearly much fainter than ESO 137-001's tails, especially given ESO 137-001's much larger distance (∼4.3 times) and the much higher absorption to the direction of ESO 137-001. We can also define strong X-ray tails as one-sided features outside of the tidal truncation radius of the galactic halo (like the tails of ESO 137-001 and ESO 137-002), in order to distinguish them from the weak X-ray tails of NGC 4438 and NGC 4388.

The observed X-ray emission of the tails is likely from the interfaces of the hot ICM and the cold stripped ISM, reflecting the multi-phase nature of mixing. Although the detailed mixing process that determines the emergent X-ray spectra is poorly understood, the observed tail temperature may be related to the mass-weighted temperature that is usually adopted in simulations, T_mw = (M_ISMT_ISM + M_ICMT_ICM)/(M_ISM + M_ICM) ≈ T_ICM/(1 + X), where X = M_ISM/M_ICM and we assume that T_ISM ≪ T_ICM. For the X-ray tails of ESO 137-001 and ESO 137-002, X = 2–6.5 if we assume T_mw is the observed spectroscopic temperature. However, it is easy to find out that the mass-weighted temperature is always larger than the spectroscopic temperature derived by fitting the observed spectra with a single-kT model (T_spe). We examined this with the CEMEKL and APEC models. For the CEMEKL model, dEM(T) ∝ (T/T_max)^α−1dT. Assuming an isobaric condition and for α> − 1,

$$\begin{equation} T_{\rm mw} = \frac{\int T dm(T)}{\int dm(T)} = \frac{1+\alpha }{2+\alpha } T_{\rm max}. \end{equation} \tag{ 1 }$$

For α> − 0.5, 1/3 T_max < T_mw < T_max. It is also reasonable to assume that T_max ⩽ T_ICM. However, with a series of XSPEC simulations, we found that T_spe < T_mw. For example, T_spe = 0.5–1 keV when −0.5 < α < 0.5, while T_mw = 2–3.6 keV for T_max = 6 keV. This is not surprising as T_spe is mainly determined by the centroid of the iron-L hump at low temperatures, which biases T_spe low. As long as there are significant emission components at kT = 0.4–2 keV, the iron-L hump will be strong to make T_spe deviated from T_mw. On the other hand, the parameter α in the CEMEKL model may reflect the age of the X-ray tail. Our simulations indeed show that T_spe increases monotonously with α. From the CEMEKL fits to the spectra of ESO 137-001 and ESO 137-002, ESO 137-002's stripping event may indeed be older. The above simple estimate implies that T_spe is a certain fraction of the ICM temperature. The luminosity of the X-ray tail should also increase with the magnitude of the ambient ICM pressure, under an isobaric condition. Thus, the X-ray tails of late-type galaxies in groups and the outskirts of poor clusters may have temperatures (e.g., <0.4 keV) and densities too low to be bright in the Chandra and XMM-Newton energy bands. This helps us to explain the paucity of strong X-ray tails of late-type galaxies in nearby clusters (e.g., Sun et al. 2007b). Other factors include the generally high cluster background and the small number of late-type galaxies in the Chandra field of nearby clusters as the covered cluster volume is small.

With the above scenario, the X-ray tail of a late-type galaxy is simply the manifestation of its cold ISM tail embedded in the hot ICM. Optical line emission is also expected and is indeed detected, e.g., the Hα tails of ESO 137-001 and ESO 137-002 (S07 and Section 6). H i observations of A3627 were made with Australia Telescope Compact Array (ATCA) in 1996, and no H i emission was detected from ESO 137-001 or ESO 137-002 (Vollmer et al. 2001b). The 3σ detection limit is ∼3 mJy beam⁻¹ in one velocity channel with a beamsize of 30'' (Vollmer et al. 2001b). Adjusting to the cluster distance used in this work and assuming a linewidth of 150 km s⁻¹ (Vollmer et al. 2001b), the corresponding H i gas mass limit is ∼5 × 10⁸ M_☉ per beam. While this limit is not weak for the remaining H i gas in the disks of ESO 137-001 and ESO 137-002 (roughly within one beam), the X-ray tails of ESO 137-001 and ESO 137-002 span over 7–15 times the beam size so a lot of diffuse H i gas can remain undetected downstream of the galaxy. In fact, with the sensitivity of the available ATCA data (an H i column density limit of 2 × 10²⁰ cm⁻²), four of the seven H i tails in the Virgo cluster found by Chung et al. (2007) and the long H i tail of NGC 4388 (Oosterloo & van Gorkom 2005) would not be detected. Clearly, deeper H i observations are required, although the nearby bright radio source PKS 1610–60 (43 Jy at the 1.4 GHz) makes the task difficult (8'–14' from ESO 137-001 and ESO 137-002). Both ESO 137-001 and ESO 137-002 are expected to host a significant amount of atomic and molecular gas initially. From Blanton & Moustakas (2009), the total neutral plus molecular gas fraction (relative to the stellar mass) is ∼70% for ESO 137-001 and ∼30% for ESO 137-002. Assuming a type of Sc and a blue-band diameter of 20 kpc, the expected H i mass of ESO 137-001 is ∼2.4 × 10⁹ M_☉, from the empirical relation derived by Gavazzi et al. (2005). Similarly, for ESO 137-002 (a type of Sa and a blue-band diameter of 30 kpc), the expected H i mass is ∼3.9 × 10⁹ M_☉. Thus, ESO 137-001 is expected to have ∼(3–5) × 10⁹ M_☉ of the cold ISM initially, while ESO 137-002 is expected to have ∼(4–9) × 10⁹ M_☉ initially. These numbers can be compared with the X-ray gas mass in their X-ray tails (Table 2). Sivanandam et al. (2010) detected the mid-IR emission from warm H₂ gas (∼150 K) in the galaxy and the first 20 kpc of the main tail, with a total mass of ∼2.5 × 10⁷ M_☉. The H₂ lines that IRS detects are rotationally excited lines that are generated very efficiently. The warm gas they probe is likely a small fraction of the total H₂ gas, if the bulk of the molecular gas is still cold. Deep H i and CO observations are required to recover the bulk of the cold gas in ESO 137-001, ESO 137-002, and their wakes.

The same question can also be asked to other cluster late-type galaxies with signs of stripping. Does a stripped ISM tail show up in H i, Hα, and X-rays simultaneously? We attempt to summarize the known examples of stripped tails for cluster late-type galaxies. A Venn diagram is plotted in Figure 15. While a lot of stripped tails are known for cluster late-type galaxies, there is currently little known overlap between different bands, especially between X-ray tails and H i tails. This may be due to the lack of sufficient data, either in X-rays (e.g., for the Virgo galaxies with H i tails; Chung et al. 2007) or in H i (e.g., the two galaxies in this paper). We are pursuing XMM-Newton data for the Chung et al. (2007) galaxies and deeper ATCA data for the A3627 galaxies to better address this question. On the other hand, there are various reasons that a one-to-one correlation may not exist in every case, even with better data. Depending on its age and the surrounding ICM, an ISM tail may emit predominantly in the H i band or the X-ray band. Soft X-ray emission of the tail may only be bright in high-pressure environment. We also understand little about the details of mixing. It may be related to the mean free path of particles and the coherence length of the magnetic field, which would introduce a radial dependence on the efficiency of mixing. Nevertheless, we first need better data to update the Venn diagram shown in Figure 15.

Zoom In Zoom Out Reset image size

Deep X-ray data alone, like ESO 137-001's, can reveal some thermal history of the stripped ISM. One surprise from ESO 137-001's X-ray data is the constancy of the spectroscopic temperatures along both tails (Figure 7). Indeed, extra care is required to interpret this spectroscopic temperature as it is mainly determined by the centroid of the combined iron-L humps from multi-T gas. Nevertheless, the possible spectral difference between ESO 137-001's tails and ESO 137-002's tail is intriguing. This difference, if confirmed, combined with the temperature constancy of ESO 137-001's tails, puts a constraint on the heating timescale of the stripped cold ISM. Clearly, the knowledge of the H i/CO gas distribution in these two galaxies will help us to tighten the constraint, although it is certain that this timescale depends on the ICM temperature.

After the initial analytic work by Gunn & Gott (1972), many simulations have been run on the ram pressure stripping of late-type galaxies in clusters (e.g., Abadi et al. 1999; Schulz & Struck 2001; Vollmer et al. 2001a; RB08; Kapferer et al. 2009; Tonnesen & Bryan 2009, and references in those papers). Recent efforts surpass the early ones in many respects, e.g., resolution, time steps, realistic treatment of the galaxy orbit, and some important baryon physics (e.g., cooling and star formation). We mainly compare our results with recent simulations of three groups, the Innsbruck group with the GADGET-2/SPH code (Kronberger et al. 2008; Kapferer et al. 2009), the Bremen group with the FLASH/AMR code (Roediger et al. 2006; RB08), and the Columbia group with the Enzo/AMR code (Tonnesen & Bryan 2009). The GADGET-2 simulations include recipes for cooling, star formation, and stellar feedback. The model galaxies run through a wind tunnel with a constant pressure. The FLASH simulations are non-radiative and model the flight of a disk galaxy through a galaxy cluster with realistic ICM and potential distributions. The Enzo simulations include cooling but not feedback (Tonnesen & Bryan 2009). To mimic effects of some heating processes not in simulations, they impose a minimum temperature floor. The model galaxy is also subject to a constant ram pressure. The inclusion of cooling is important, as it indeed happens as revealed from our data (the intracluster H ii regions and star clusters) and it has profound effects in simulations (Kapferer et al. 2009; Tonnesen & Bryan 2009). Cooling also naturally adjusts the ISM distribution on the disk plane, creating multi-phase ISM with enhancement and holes (Kapferer et al. 2009; Tonnesen & Bryan 2009), which is not present in the non-radiative runs. However, cooling needs to be offsetted by stellar feedback. Unknown efficiencies of transport processes (viscosity and heat conduction) because of unclear strength and configuration of magnetic field, plus the other uncertainties in the prescriptions of star formation and stellar feedback, make the tasks difficult (e.g., Tonnesen & Bryan 2009). We emphasize that comparison at the current stage is not straightforward as galaxies in these simulations are always more massive than ESO 137-001, and the high-pressure environment was not explored in all simulations. We also caution that it is best to compare the observational data with the mock data generated from simulations. Kapferer et al. (2009) and Tonnesen & Bryan (2009) attempted to create X-ray mock data but more work is required, e.g., folding with the actual Chandra response, projection on the ICM background, and the use of spectroscopic temperature rather than mass-weighted temperature (Section 7.1).

Tails generally appear longer and wider in simulations than in observations. The simulations by Kapferer et al. (2009) show that the X-ray tails can be traced to ∼300 kpc from a galaxy that is about four times more massive than ESO 137-001, with an ambient pressure similar to ESO 137-001's. The simulated X-ray tails are also very clumpy, with widths generally increasing with the distance from the galaxy. The FLASH simulations also produce longer tails than observed in this work, but are difficult to compare with the X-ray data directly. As RB08 pointed out, the observed tail length depends on the mass loss per orbital length, which is the highest in poor clusters as galaxies in rich clusters move too fast and the ram pressure in groups is not high enough. The FLASH simulations also show tail flaring (i.e., increasing width with distance from the galaxy downstream) in all cases. The velocity perpendicular to the direction of the ram pressure comes from the random motion of the ICM, the disk rotation and turbulence in the wake. For a rotation velocity of 100 km s⁻¹ over 50 Myr, the offset is 5 kpc, which is not small in ESO 137-001's case as its tails are narrow. The Tonnesen & Bryan (2009) simulations produce long but narrower tails. The tails are also very clumpy. None of these simulations can reproduce the double tails of ESO 137-001. The simulated tails are generally too clumpy with strong turbulence. This simple comparison may point to a higher ICM viscosity than present in the simulations. The Reynolds number is ${\sim} 3{\cal M}(L/\lambda)$, where ${\cal M}$ is the Mach number, L is the size of the remaining galactic gas disk, and λ is the mean free path of particles in the ICM. For unmagnetized gas, λ∼ 10 kpc around ESO 137-001. The radius of the remaining Hα disk and the galactic X-ray emission is only ∼1.7 kpc (S07). Thus, the Reynolds number is on the order of unity which implies a laminar flow. Of course, the presence of magnetic field would increase the Reynolds number. In the simulations by Kapferer et al. (2009), a bright spot in the X-ray images is always found at distances of a few tens of kpc from the galaxy, caused by compressional heating. This feature can be compared with the bright spot in the main tail (Figures 1 and 5). However, the location of this feature is generally too far from the galaxy in simulations with high ambient pressure. Such a bright spot is also not observed in the secondary tail.

The most obvious question about ESO 137-001's X-ray tails is why are there double tails. The simplest explanation is to involve another dwarf galaxy forming a pair with ESO 137-001. This was in fact simulated by Kapferer et al. (2008), showing double gaseous tails, stellar tidal tails, and a stellar bridge between two galaxies. The X-ray gas mass of the secondary tail is 40%–50% of the mass of the main tail, assuming the same filling factors of the X-ray gas. If we simply take the ratio of their X-ray gas masses as the ratio of their ISM gas masses, the stellar mass of this assumed dwarf that is responsible for the secondary tail should be ∼30% of ESO 137-001's stellar mass, from the ISM mass–stellar mass relation summarized by Blanton & Moustakas (2009). Such a galaxy should easily be detected, at least through the tidal tails and bridge it causes even if it hides behind ESO 137-001. However, the SOAR images do not show the presence of another galaxy near ESO 137-001. There are also no tidal features upstream of the galaxy. There are some continuum features downstream of the galaxy as shown in S07. However, all of them are around confirmed H ii regions and have blue colors. Our new HST data only reinforce the conclusion from the SOAR images. The central region of the galaxy is very dusty but the galaxy appears regular without significant tidal features in the HST I band (M. Sun et al. 2010, in preparation). Thus, we conclude that both X-ray tails originate from ESO 137-001.

The discovery of two separated stripping tails from one galaxy presents a challenge to our understanding of stripping. In simulations (e.g., RB08; Kapferer et al. 2009; Tonnesen & Bryan 2009), the cold ISM is removed from the sides of the disk or through holes, and is quickly mixed to form one clumpy tail because of disk rotation, momentum transferred from the ICM, and turbulence in the wake, especially if projected on the plane of sky. Perhaps we should begin to consider more realistic ISM distributions in the galaxy. Optical images clearly show the presence of two thick spiral arms, one to the north and the other one to the south (S07; better shown in the HST images). Maybe the two X-ray tails correspond to stripping of molecular gas concentrated around these two spiral arms. CO images of nearby face-on spiral galaxies indeed show that molecular gas distribution follows major spiral arms in many cases (e.g., Kuno et al. 2007). This is also generally true for H i gas (e.g., Adler & Westpfahl 1996). For ESO 137-001, we present a simple cartoon in Figure 16. The north arm is apparently the trailing arm where stripping is easier so a brighter X-ray tail is produced. The south arm is the leading arm so gas removal there is more difficult as most gas would be pushed inward to deeper galactic potential first. This also explains the fact that more intracluster H ii regions (at >15 kpc from the nucleus) are found around the main tail while the numbers of H ii regions in the halo are similar at two sides. The curvature of the secondary tail may come from the galactic rotation, which is high enough to cause the observed offset. The main tail may also be curved in three-dimension but appears straight in projection. The rather regular wake structure may imply significant viscosity. On the other hand, draping of the ICM magnetic field (e.g., Dursi & Pfrommer 2008) can also inhibit the motion of the stripped ISM perpendicular to the direction of ram pressure.

Zoom In Zoom Out Reset image size

In summary, it remains to be seen whether simulations can reproduce the double X-ray tails of ESO 137-001 as observed, and the gas properties in other bands (Hα and H₂ at least). We also emphasize that the comparison between data and simulations is not trivial and straightforward. Observationally, we need deeper data at multiple wavelengths. For simulations, mock data need to be produced with sensitivities matching the actual observations. Uncertainties in the viscosity, heat conduction, cooling, and stellar feedback should be explored with different runs. Eventually, we need to compare the multi-wavelength data (H i, CO, IR, optical, and X-rays) with the mock data in the same band. Besides the simple comparison of morphology, a lot of details of the tail properties (e.g., temperature distribution, filling factors of the gas in different phases, and energy transfer between gas in different phases) can be better examined.

The new observations of ESO 137-001, for the first time, unambiguously confirm the active star formation in the cold ISM stripped by the ICM ram pressure. As emphasized in S07, ESO 137-001's immediate surroundings are devoid of bright galaxies. There are only two galaxies that are within 4 mag of ESO 137-001 in the I band, within 100 kpc from the end of ESO 137-001's X-ray tail (G1 and G2 in Figure 1 of S07). Their redshifts are unknown and they are 1.9–2.7 mag fainter than ESO 137-001 in the I band. There is also no such galaxy within 70 kpc from the nucleus of ESO 137-001. Both G1 and G2 are undisturbed and there are no tidal features between ESO 137-001 and G1 (or G2). Our HST data do not reveal another dwarf galaxy near ESO 137-001 (M. Sun et al. 2010, in preparation). The velocity map of the H ii regions also shows no extra component, besides ESO 137-001's rotation pattern. It is also known that the cluster potential is not the main factor for stripping (S07; Sivanandam et al. 2010), although it indeed affects the size of ESO 137-001's dark matter halo. Thus, we conclude that ICM ram pressure is responsible for the stripping of the ISM. In our case, the ambient pressure is even high enough to remove some molecular gas from the galaxy. The H i gas has typical pressures of P/k_B ∼ 10³–10⁴ K cm⁻³, while the pressure in the core of a starless giant molecular cloud can be higher than 10⁶ K cm⁻³. The ambient total thermal pressure (electron + ions) is 1.3 × 10⁵ K cm⁻³ for n_e = 10⁻³ cm⁻³ and kT = 6 keV, while the ram pressure is 3.2 × 10⁵ K cm⁻³ for n_e = 10⁻³ cm⁻³ and v_gal = 1500 km s⁻¹.

After removing the cold ISM from the galaxy, the key for the intracluster star formation is that the stripped ISM must be able to cool, likely aided by shocks and the absence of the UV radiation field that is strong on the disk plane. Intracluster star formation in the stripped ISM was first suggested in simulations (Schulz & Struck 2001; Vollmer et al. 2001a). Schulz & Struck (2001) even suggested formation of dwarf galaxies in this mode. In the new GADGET-2/SPH simulations by Kronberger et al. (2008) and Kapferer et al. (2009), new stars are formed in the stripped ISM, forming ∼1 kpc structures with masses of up to 10⁷ M_☉ individually to hundreds of kpc downstream of the galaxy. However, all simulations suffer from uncertainties in transport processes like viscosity and heat conduction, which are important for determining the heating rate.

More evidence of intracluster star formation, either H ii regions or blue star clusters, has been revealed recently (e.g., Lee et al. 2000; Gerhard et al. 2002; Cortese et al. 2004, 2007; Yoshida et al. 2008). They are only at one side of the galaxy. In some cases, tidal interaction may be important (e.g., Lee et al. 2000). But they all lack unambiguous associations with gaseous tails. We especially highlight the growing number of examples of one-sided blue stellar filaments and star clusters behind cluster galaxies, which includes two galaxies in A2667 and A1689 at z∼ 0.2 (Cortese et al. 2007) and a dwarf galaxy RB 199 in the Coma cluster (Yoshida et al. 2008). The connection between the first two galaxies at z ∼ 0.2 with ESO 137-001 was discussed in S07. In the case of RB 199, Yoshida et al. (2008) found narrow blue filaments, knots, Hα filaments, and clouds extending to 80 kpc from the galaxy in one side. It is a rather remarkable case as the galaxy is much fainter than all previous galaxies, with a stellar mass of ∼8 × 10⁸ M_☉. The stellar mass of the downstream structures is ∼10⁸ M_☉, which is ∼10% of the galactic mass. Yoshida et al. (2008) also favored a mechanism similar to what we suggested for ESO 137-001 in S07. These three examples may imply a rather common evolution stage for cluster late-type galaxies. As the stage with bright blue stellar trails is short (⩽200 Myr), more similar examples are expected only in large surveys.

If we assume that intracluster star formation happens in the stripped ISM of every cluster spiral, can this mode contribute a significant fraction of the intracluster light? The answer depends on the mass fraction of intracluster stars formed out of the stripped ISM (or the star formation efficiency). S07 gave a rough estimate of ∼1%, which may only be a lower estimate as blue stellar filaments and clusters (not shining in Hα) are not counted. While a detailed account of the total stellar mass downstream of ESO 137-001 is the aim of M. Sun et al. (2010, in preparation), here we take 1% as a lower limit. If the efficiency of the intracluster star formation is only 1%, the intracluster light contributed from star formation in stripped ISM is small, given the mass fraction of the intracluster light to the total stellar light in clusters (∼20%, likely depending on the halo mass; e.g., Krick & Bernstein 2007). However, the case of RB 199 implies a fraction of ∼10%, if the original ISM in RB 199 had a mass comparable to its stellar mass (e.g., Blanton & Moustakas 2009). In the simulations by Kronberger et al. (2008) and Kapferer et al. (2009), 7%–12% of the initial ISM turns into new stars in the wake. If an efficiency of 10% is adopted, the intracluster light contributed by ram pressure stripping is a significant fraction of the total intracluster light. However, this efficiency may not be too high given the star formation efficiency in galaxies. In galactic disks, ∼1% of gas is converted to stars per free-fall time (e.g., Leroy et al. 2008). In 100 Myr (a typical orbital time in the disks), ∼6% of gas is converted to stars (e.g., Kennicutt 1998; Leroy et al. 2008). We note that 100 Myr is also about the orbital time of the residual Hα disk (∼1.7 kpc radius with an orbital velocity of ∼100 km s⁻¹) and close to the age of the X-ray tails. Clearly, a lot of work is required to constrain the intracluster star formation efficiency. The efficiency may also change with the mass of the host galaxy as more stars are formed inside the halo of more massive galaxies. In principle, the ratio of the intracluster Type Ia supernova to the other types can also inform us of the significance of intracluster star formation.

As active star formation is confirmed downstream of the galaxy, the discovery of ULXs is not a surprise. As we estimated in Section 5.2, the total luminosity of ULXs can be consistent with the expectation from the intracluster SFR. The space density of X-ray point sources also traces the strength of the star formation, stronger in the halo and weaker around the unbound tail. A similar system may be the high-velocity system of NGC 1275. Gonzalez-Martin et al. (2006) found a concentration of eight ULXs (L_{0.5–7 keV} > 3 × 10³⁹ erg s⁻¹) around the high-velocity system. They may be associated with the active star formation in and downstream of the high-velocity system, similar to what is happening in ESO 137-001.