(1992), with a lensing galaxy at a redshift of z \ 0.647

Transcription

1 THE ASTRONOMICAL JOURNAL, 5:È5, 998 January ( 998. The American Astronomical Society. All rights reserved. Printed in U.S.A. REDSHIFTS OF THE GRAVITATIONAL LENSES B422]23 AND PG 5]8 JOHN L. TONRY Institute for Astronomy, University of Hawaii, Honolulu, HI 96822; jt=avidya.ifa.hawaii.edu Received 997 June 9; revised 997 September 24 ABSTRACT B422]23 and PG 5]8 are gravitational lens systems that produce quadruple QSO images. In each, there is real promise that time delays can constrain the Hubble constant, and both belong to systems that can play an important role in modeling the lens potential. This article reports redshifts for the lensing galaxies and three neighboring galaxies in each of the two systems. B422]23 comprises a group at z \.339 with a dispersion of 733 km s~, and PG 5]8 comprises a group at z \.3 with a dispersion of 326 km s~. One of the neighboring galaxies in the B422]23 system turned out to be an emission-line galaxy at z \.536, suggesting that QSO light passing through B422]23 may have been subjected to lensing by a cluster at this more distant redshift. The velocity dispersion of the lensing galaxy in PG 5]8 is determined to be 28 ^ 25 km s~ (A square aperture), which is surprisingly large, given the image splittings of A.2 in that system. Key words: distance scale È gravitational lensing. INTRODUCTION The di ering paths followed by light around a gravitational lens produce di erent times of Ñight. If the lensed source should vary, measurement of a time delay can transform dimensionless redshifts into physical distances, providing a Hubble constant (Refsdal 964). To within factors of order unity, the entire time of Ñight is simply the time delay divided by the square of the image splitting (in radians). In principle, this is an extremely promising method for learning about cosmology; in practice, it has proved difficult to measure time delays and model lenses accurately enough to challenge the accuracy of %È2% claimed by more traditional measures of H. Nevertheless, gravitational lens measurements of cosmology are critically important, because they are potentially subject to far fewer biases than the usual distance ladder. Time delays have now been measured for two lens systems, 957]56 and PG 5]8. The difficulty in measuring a time delay and then rendering cosmological information from it is illustrated by 957]56. A time delay was reported by Schild & CholÐn (986), disputed by Press, Rybicki, & Hewitt (992), and Ðnally resolved in favor of SchildÏs number through further observations by Kundic et al. (995). Models of the system have also undergone improvement, with estimated values for H as low as 5 km s~ Mpc~, and as high as 9 km s~ Mpc~ (Grogin & Narayan 996a, 996b). PG 5]8 has been observed by Schechter et al. (997), who derived a time delay and estimated a Hubble constant of 45 km s~ Mpc~ based on image and lens positions from Kristian et al. (993) and on a lens redshift of about z B.3 from Henry & Heasley (986) and Angonin- Willaime, Hammer, & Rigaut (993). The observations reported here were recently anticipated by Kundic et al. (997a, hereafter KCBL), who Ðnd z \.3 and use this to derive H \ 52 ^ 4 km s~ Mpc~. B422]23 is a very thoroughly studied QSO and Lya system at a redshift of 3.62 discovered by Patnaik et al. Based on observations made at the W. M. Keck Observatory, which is operated jointly by the California Institute of Technology and the University of California. (992), with a lensing galaxy at a redshift of z \.647 according to Hammer et al. (995). There is a small group of three galaxies of unknown redshift within a few arcseconds. Although no time delay has been measured in this system, the photometry compiled by Keeton & Kochanek (996) suggests that there may be enough variability in the QSO that it should be possible to do so. This article reports the Ðrst results of an ongoing program to measure the most basic information necessary to exploit gravitational lenses: the redshift of the lensing galaxy, redshifts of galaxies clustered around the lens, and velocity dispersions of the stars within the galaxies. Although a small ingredient compared with measuring time delays and modeling potentials, it is an essential one. 2. OBSERVATIONS AND REDUCTIONS B422]23 and PG 5]8 were observed on 997 March 3 and 3, using the Low Resolution Imaging Spectrometer (LRIS; Oke et al. 995) at the Keck II Telescope on Mauna Kea. Calibration exposures and observations of the clusters MS 358]62 (to act as a velocity dispersion calibrator) and HD and AGK2 ]4 873 (as radial velocity templates) were also made. The observations are summarized in Table. The sky was clear and the seeing was about A.8 throughout both nights. Long slits of A. and A.7 were used, along with gratings of 3 and 6 lines mm~, both blazed at 5 A. The 3 line mm~ grating was always rotated to provide coverage from 38 to 87 A, whereas the 6 line mm~ grating was rotated to cover either 48È768 A for the QSO observations or 38È68 A for the templates. The spectral resolution was about 7.9 A FWHM for the 3 line mm~ grating, 4.65 A FWHM for the 6 line mm~ grating with a A. slit, and 3.59 A FWHM for the 6 line mm~ grating with a A.7 slit; the scale along the slit was A.2 pixel~. The template stars were guided smoothly across the slit in many locations in order to have uniform illumination across the slit, to build up the signal-to-noise ratio (S/N), and to map out loci of constant slit position across the detector. The slit was rotated as illustrated in Figure, either to observe two companion galaxies simultaneously or else to cover the lensing galaxy (denoted GL ÏÏ) while avoiding as much QSO light as possible.

2 2 TONRY Vol. 5 TABLE OBSERVATION LOG P.A.a Exposure Slit Width UT Object sec z (deg) (s) (arcsec) Gratingb 997 Mar 3: AGK2 ] / PG 5 GL ] B / PG 5 GL ] B / PG 5 A ] A /5... PG 5 G ] / PG 5 G ] / MS 358 G / MS 358 G / B422 ] / B422 ] / B422 GL / B422 GL / HD / HD /5 997 Mar 3: AGK2 ] / AGK2 ] / PG 5 GL ] B / PG 5 GL ] B / PG 5 A ] A /5 a East of north. b Lines per millimeter/blaze (A ). The spectra were reduced using software described in detail by Tonry (984). The basic procedure is to Ñatten the images, remove cosmic rays, derive a wavelength solution as a function of both row and column using skylines (wavelengths tabulated by Osterbrock et al. 996), derive a A2 A C GL B B Gx G4 " Slit 35 " A GL B C D Slit 67 Slit 42 Slit 33 Gx PG5+8 " G4 G MS FIG..ÈSlit positions and galaxy identiðcations for B422]23, PG 5]8, and MS 358]62. North is up and east is to the left in each panel. G slit position solution as a function of both row and column (using the positions of the template star images in the slit), rebin the entire image to coordinates of log wavelength and slit position, add images, and then sky-subtract. A quadratic Ðt to patches of sky on either side of the object (including a patch between for the galaxy pair observations) did a very good job of removing the skylines from the spectra. As has been stressed by Kelson et al. (997b), measuring the velocity dispersion of a galaxy at a redshift signiðcantly greater than zero must be done carefully, since the instrumental resolution of a spectrometer tends to be a constant increment in wavelength, whereas the redshifted spectrum has been stretched. Hence a straightforward crosscorrelation or Fourier quotient will underestimate the dispersion of the galaxy. Kelson et al. measured dispersions in the cluster MS 358]62 using high-resolution template spectra and very careful modeling of the spectrometer resolution (derived by measuring skyline widths). The observations here exploit the fact that the ratio between a A.7 and a A. slit is slightly larger than ] z for these lenses, and hence a template measured with the A.7 slit will have almost exactly the same instrumental resolution as a galaxy at a redshift of z B.3. This is borne out by the ratio of the measured spectral resolutions: (4.65 A )/(3.59 A ) \.3. This is applicable only to the 6 line mm~ observations, of course. As a test that this observing procedure will produce correct dispersions, we also observed the cluster MS 358]62, for which Kelson et al. (997a) measured dispersions of 33 ^ 6kms~ for the central galaxy (G) and 29 ^ 4kms~ for the neighboring bright galaxy (). The analysis of the companion galaxies was straightforward, since the spectra had very high S/N and were uncontaminated by QSO light. In each case the spectrum was extracted and cross-correlated with the template spectrum according to Tonry & Davis (979), as well as analyzed by the Fourier quotient method of Sargent et al. (977). (The cross-correlation is more robust in the case of low S/N, but

3 No., 998 REDSHIFTS OF GRAVITATIONAL LENSES 3 at the signal levels here the two results are statistically the same, and are simply averaged.) For each spectrum the redshift, error, and velocity dispersion were calculated. The redshift calculation included the entire spectrum, whereas the spectrum blueward of 4 A (rest frame) was excised for the dispersion calculation, since the calcium H and K lines are so broad that they are always problematic for dispersions. Dispersions are not shown for 3 line mm~ observations, and they are not corrected for any aperture e ects (hence they correspond to an aperture of approximately A square). Table 2 lists the redshifts, velocity dispersions, and crosscorrelation signiðcance r ÏÏ values for each spectrum. 2.. B422]23 We were surprised to Ðnd an unmistakable emission-line system adjacent to B422-, o set by about A. along the line toward B422-. It showed very clear, extended j3727 and j57 emission lines in both exposures, and Hb could also be faintly discerned. Lacking any imaging information on this galaxy, we tabulate it as B422-Gx.ÏÏ The lensing-galaxy extractions require considerable care. In the case of B422]23, we exploited the fact that the A/B/C components of the QSO would put a more extended distribution of light along the slit than the lensing galaxy. We therefore extracted three swaths of approximately A; the central one we termed QSO ] galaxy,ïï and the two Ñanking swaths we added and termed QSO.ÏÏ Of course, there is considerable galaxy light in the Ñanking swaths, but the expectation was that the proportion of galaxy light would be less. We found a linear function of wavelength that, when multiplied by the QSO ÏÏ spectrum, would match the Lya and C IV QSO lines in the QSO ] galaxy ÏÏ spectrum. Subtraction left behind a residue that was a clean enough lensing galaxy spectrum to allow derivation of a solid cross-correlation redshift. In principle, it would have been better to have a pure QSO spectrum for subtraction, and indeed, we obtained a pure QSO spectrum by shifting the slit down on top of the A/C line. However, the QSO light in the composite spectrum has a large gradient across the slit and, hence, a wavelength shift and large intensity variations. Time did not permit rotating the spectrometer 8 and o setting to the other side of the QSO to try to duplicate the QSO contribution to the composite spectrum. We see no emission in the B422]23 lensing galaxy whatsoever, despite the claims by Hammer et al. (995) that j3727 and j57 appeared at 638 and 8247 A. We excised the A and B bands at 7625 and 6885 A from the spectrum, and 3 A centered on the Lya peak, and performed a cross-correlation against the template star. The crosscorrelation came out with a correlation peak at z \.3366, with an r-value of 4.2, which is signiðcant enough that it is quite unlikely (probability D%) to be spurious. It is possible to see Ca II H ] K, Mg, and Na at that redshift. With the corroboration from the neighboring galaxies, we are quite conðdent that this is indeed the redshift of the lensing galaxy. The spectra of the QSO, lensing galaxy,,, and Gx are plotted in Figure 2. The most prominent Fraunhofer lines are labeled for the lensing galaxy, and while they can plausibly be seen by eye, our faith in this redshift ultimately depends on the cross-correlation signiðcance PG 5]8 As with B422]23, we serendipitously picked up an emission-line galaxy in the PG 5]8 system, along the slit between G and, lying 3A.8 from. It displays j57, j4959, Hb, and Hc emission lines, and we refer to it in Table 2 as PG 5-Gx.ÏÏ The slit for PG 5]8 ran over the top of the QSOÏs B component (the lensing galaxy) and skirted the edge of the C component, creating a bimodal distribution of light in the slit with the galaxy in the center. In this case, the QSO light illuminated the slit uniformly enough that the subsequent exposure centered on the A-A2 components matched the QSO light from the B and C components to a high degree of accuracy. The B and C peaks were separated by A.8 at this position angle, and we extracted the inner A from between the two peaks and subtracted a scaled version of the A-A2 spectrum to obtain the lensing-galaxy spectrum illustrated in Figure 3. Comparison with the redshifts presented by KCBL shows extremely good agreement: GLÈ.3 KCBL,.398 here; GÈ.399 KCBL,.398 here; È.32 KCBL,.323 here. As noted by KCBL, these redshifts are in good agreement with the results of Henry & Heasley (986) but are only marginally consistent with those of Angonin-Willaime et al. (993). The spectra of the QSO, lensing galaxy, G,, and Gx are plotted in Figure 3. TABLE 2 REDSHIFTS AND DISPERSIONS Slit Position Dispersion Galaxy (arcsec) z (km s~) ra B422-GL ^ B [ ^ B ^ B422-Gx... [ ^.5... em. PG 5-GL ^.2 28 ^ 25.4 PG 5-G... [ ^. 256 ^ PG ^. 3 ^ 6.9 PG 5-Gx ^.2... em. PG 5-GL ^ MS 358-G ^. 299 ^ 22.9 MS [ ^. 235 ^ MS [ ^. 55 ^ MS 358-G ^.2 75 ^ a SigniÐcance of cross-correlation with template spectrum; em.ïï denotes emission.

4 4 TONRY Vol. 5 FIG. 2.ÈSpectra for (o set by 5 counts for clarity),, Gx, QSO ] lensing galaxy, pure ÏÏ QSO, and extracted lensing galaxy (scaled by a factor of for visibility) in the B422]23 system. The spectra have been Gaussian smoothed with an FWHM of 6 A. The top axis shows rest wavelength at a redshift of.34. The [O II] and [O III] lines in Gx are labeled, as are Fraunhofer lines in the lensing galaxy ( GL ÏÏ) MS 358]62 A pair of exposures of MS 358]62 identical to those of PG 5]8 yielded spectra of seven galaxies. The slit was aligned between the central galaxy (called G here) and the bright galaxy 9A to the southwest (P.A. \ 24 ) (called ). Galaxies and G4 were well centered in the slit, and hence their velocities and dispersions should be reliable. The three remaining galaxies were marginally caught by the slit and were of low S/N, so we do not report their velocities. Comparison with Fabricant, McClintock, & Bautz (99) for the three galaxies in common (G \ their No. 4, \ No. 2, and \ No. 26) yields a mean o set of dz \.2 and a scatter of p \., well within the error estimates. z The comparison of velocity dispersions with those of Kelson et al. (997a) reveals no signiðcant o set in the dispersions that we measure: GÈ33 ^ 5kms~, 299 ^ 22 here; È29 ^ 4kms~, 235 ^ 23 here. The mean of these ratios di ers from unity by only 3%, indicating that the dispersions and errors reported here should be accurate. 3. DISCUSSION Computing the means and standard deviations for the redshifts presented here, we Ðnd that the B422]23 lens is part of a group at z \.339 ^.2 with a rest-frame dis- g persion (i.e., standard deviation of cz divided by ] z) of 733 km s~ (three galaxies). The PG 5]8 lens is part FIG. 3.ÈSpectra for G,, Gx, QSO ] lensing galaxy, pure QSO, and extracted lensing galaxy in the PG 5]8 system. The spectra have been Gaussian smoothed with an FWHM of 6 A. The top axis shows rest wavelength at a redshift of.3. Emission lines of j57, j4959, and j486 are apparent in the spectrum of Gx. of a group at z \.3 ^. with a rest-frame disper- g sion of 326 km s~ (Ðve galaxies), consistent with the result from KCBL of 27 ^ 7 km s~. Using the four galaxies observed in the MS 358]62 cluster, we Ðnd z \.325 c ^. with a rest-frame dispersion of 432 km s~ (four galaxies). A few comments seem appropriate. The B422]23 group appears to be quite massive, particularly given the number of galaxies present. Deeper imaging would probably be worthwhile. Although the group dispersion is based on only three galaxies and is therefore quite uncertain, there is no reason to exclude galaxy from the dispersion calculation, despite the fact that its velocity di ers from G by more than 3 km s~. For example, moving away from M87 in the Virgo Cluster, the Ðrst bright galaxy encountered is NGC 446, which has a relative velocity of about 5 km s~. B422]23 could be quite comparable to Virgo in velocity dispersion, although its compactness would suggest a smaller cluster mass. KCBL compared the PG 5]8 cluster to a Hickson poor group (Hickson et al. 992), but the dispersion here of 326 km s~ is at the very high end of such groups and is in quite good agreement with the models of Schechter et al. (997), who needed a group dispersion of 383 km s~ to provide the necessary shear to explain the image positions. Finally, the dispersion for MS 358]62 serves as a warning that small numbers of galaxies near the centers of groups may not be a reliable measure of the overall group

5 No., 998 REDSHIFTS OF GRAVITATIONAL LENSES 5 mass. Fabricant et al. (99) determined a dispersion of 25 km s~ for MS 358]62 from approximately 65 galaxies within 2@ of the brightest cluster galaxy (BCG). Recomputing a dispersion from their redshifts, but restricting the sample to galaxies closer to the BCG than, we Ðnd a dispersion of 625 km s~, demonstrating how a cool subsystem can exist in a massive cluster. The redshift found here for the B422]23 system is signiðcantly lower than that reported by Hammer et al. (995), so the lens and companions are not as highly luminous as has been suggested. This lower redshift is in quite good agreement with the redshift estimated photometrically by Impey et al. (996). The rather high cluster mass called for by Hogg & Blandford (994) appears to be conðrmed here. The presence of the background galaxy at z \.536 at a distance of 9A from the lens may point to a background cluster that is providing signiðcant multiple lensing. Very deep imaging of this source would help elucidate whether there is indeed a background cluster present. The 28 km s~ dispersion for the lensing galaxy in PG 5]8 can be corrected according to the formula given by JÔrgensen, Franx, & Kj~rgaard (995) to the standard metric aperture commonly used in fundamental-plane studies: 3A.4 at the distance of Coma. This corrected dispersion is 293 km s~, which seems surprisingly large. Perhaps the best way to see this is to consider the Einstein ring created by an isothermal lens of dispersion p aligned with the source. This ring will have a radius r given by r \ D ds D 4np2 d. () D c2 s Using redshifts of z \.722 and z \.3 and ) \, we s d Ðnd that the Ðrst term is 394 h~ Mpc and the angular-size distance of the lens is D \ 58 h~ Mpc (i.e., B25 times the d distance of Virgo), so that A is 2.8 h~ kpc. This then yields r \ 4.72 h~ kpc \ A.68. However, the geometric mean of the four QSO positions around the lensing galaxy is A.7, which is the reason that the models of Schechter et al. favor a dispersion for the lens of about 235 km s~. The position of the ring of images is quite robust to the details of shear and azimuthal image position, so it will take detailed modeling to understand how this high dispersion can Ðt in with the rest of the observables. It is conceivable that this dispersion is simply wrong, but the spectrum is of quite high quality, even given the necessity for QSO subtraction, having an S/N of about 2 A ~. It is also possible that the dispersion drops rapidly from the measured value of 28 km s~ to something nearer 235 km s~ at a radius of 3.3 h~ kpc, where the images lie. While not unprecedented, this would be a surprisingly fast decline in velocity dispersion. Although the prospects for improved spatial resolution in the dispersion are not good, this value can be used to constrain possible models for the galaxy: an isothermal mass distribution seems problematic. The models of Keeton & Kochanek (997) explore some of these possibilities. A Ðnal possibility is that the shear from the cluster is pushing the images closer together, overcoming some of the splitting from the lens itself, but it is unclear whether such a model could be tenable. In recent years, the sophistication of gravitational lens models has clearly outstripped the quality of the data available. The Keck telescopes and the LRIS are truly marvelous facilities, and it is the aim of this and subsequent papers to try to bring some grist to this hungry mill, and to perhaps help us learn about the cosmology of our universe. Note added in manuscript.èkundic et al. (997b) have once again independently obtained results for B422]23; they Ðnd the same redshift, making it virtually certain that the lens shares the redshift of the surrounding group. They obtained a somewhat lower dispersion of 55 km s~ for the group, which is probably more accurate, since it is based on more redshifts. Thanks are due to Paul Schechter for encouragement and helpful suggestions throughout this project. REFERENCES Angonin-Willaime, M.-C., Hammer, F., & Rigaut, F. 993, in Gravita- Kristian, J., Groth, E. J., Shaya, E. J., Schneider, D. P., & Holtzman, J. A. tional Lenses in the Unverse, ed. J. Surdej, D. Fraipont-Caro, E. Gosset, 993, AJ, 6, 33 S. Refsdal, & M. Remy (Lie` ge: Inst. dïastrophys.), 85 Kundic, T., Cohen, J. G., Blandford, R. D., & Lubin, L. M. 997a, AJ, 4, Fabricant, D. G., McClintock, J. E., & Bautz, M. W. 99, ApJ, 38, (KCBL) Grogin, N. A., & Narayan, R. 996a, ApJ, 464, 92 Kundic, T., Colley, W. N., Gott, J. R., Malhotra, S., Pen, U., Rhoads, J. E., ÈÈÈ. 996b, ApJ, 473, 57 Stanek, K. Z., & Turner, E. L. 995, ApJ, 455, L5 Hammer, F., Rigaut, F., Angonin-Willaime, M.-C., & Vanderriest, C. 995, Kundic, T., Hogg, D. W., Blandford, R. D., Cohen, J. G., Lubin, L. M., & A&A, 298, 737 Larkin, J. E. 997b, AJ, 4, 2276 Henry, J. P., & Heasley, J. N. 986, Nature, 32, 39 Oke, J. B., et al. 995, PASP, 7, 375 Hickson, P., Mendes de Oliveira, C., Huchra, J. P., & Palumbo, G. G. C. Osterbrock, D. E., Fulbright, J. P., Martel, A. R., Keane, M. J., & Trager, 992, ApJ, 399, 353 S. C. 996, PASP, 8, 277 Hogg, D. W., & Blandford, R. D. 994, MNRAS, 268, 889 Patnaik, A. R., Browne, I. W. A., Walsh, D., Cha ee, F. H., & Foltz, C. B. Impey, C. D., Foltz, C. B., Petry, C. E., Browne, I. W. A., & Patnaik, A. R. 992, MNRAS, 259, P 996, ApJ, 462, L53 Press, W. H., Rybicki, G. B., & Hewitt, J. N. 992, ApJ, 385, 44 JÔrgensen, I., Franx, M., & Kj~rgaard, P. 995, MNRAS, 276, 34 Refsdal, S. 964, MNRAS, 28, 37 Keeton, C. R., & Kochanek, C. S. 996, in IAU Symp. 73, Astrophysical Sargent, W. L. W., Schechter, P. L., Boksenberg, A., & Shortridge, K. 977, Applications of Gravitational Lensing, ed. C. S. Kochanek & J. N. ApJ, 22, 326 Hewitt (Dordrecht: Reidel), 49 Schechter, P. L., et al. 997, ApJ, 475, L85 ÈÈÈ. 997, ApJ, 487, 42 Schild, R. E., & CholÐn, B. 986, ApJ, 3, 29 Kelson, D. D., Illingworth, G. D., Franx, M., & van Dokkum, P. G. 997a, Tonry, J. L. 984, ApJ, 279, 3 in preparation Tonry, J. L., & Davis, M. 979, AJ, 84, 5 Kelson, D. D., van Dokkum, P. G., Franx, M., Illingworth, G. D., & Fabricant, D. 997b, ApJ, 478, L3