X-ray images of the extragalactic sky routinely taken with Chandra and XMM-Newton do not typically detect normal galaxies as serendipitous sources in the field. Instead the images reveal a relativity sparse population of point sources, the majority of which are Active Galatic Nuclei (AGN) with a space density of order a thousand per square degree. Normal galaxies are not detected because the X-ray luminosity of normal galaxies is relatively low and the predicted fluxes very faint. However, in the deepest few million second or more exposures made with Chandra (the Chandra Deep Fields - CDFs; Giacconi et al. 2002, Alexander et al. 2003) faint X-ray emission has been detected from optically bright galaxies at redshifts of 0.1 to 0.5 (Hornschemeier et al. 2001). These are amongst the faintest X-ray sources in the CDF, with fluxes of ~ 10-16 erg cm-2 s-1 , corresponding to a luminosity of 1039 to 1041 erg s-1 - the range seen from nearby galaxies (e.g. see Shapley, Fabbiano & Eskridge 2001). Some of these might be galaxies containing a low luminosity AGN, but most are likely to be part of an emerging population of normal galaxies at faint X-ray fluxes.
The detection sensitivity of Chandra can be increased by `stacking' analysis, i. e. by `stacking' sub-images centered on the positions of galaxies in comparable redshift ranges. This can push the threshold of Chandra to ~ 10-18 erg cm-2 s-1 - equivalent to an effective exposure time of several months or more. Brandt et al. (2001) used this technique for 24 Lyman Break galaxies at z ~ 3 in the Hubble Deep Field North (Steidel et al. 1996) and detected a signal with an average luminosity of 3 × 1041 erg s-1 - similar to that of nearby starburst galaxies. Nandra et al. (2002) confirmed this result by increasing the number of Lyman Break galaxies to 144 and then extended it to also include 95 Balmer Break galaxies at z ~ 1. The Balmer Break galaxies were detected with a lower average luminosity of 7 × 1040 erg s-1 , but a similar X-ray to optical luminosity ratio as the Lyman Break galaxies. Hornschemeier et al. (2002) report `stacking' detections of optically luminous spiral galaxies at 0.4 < z < 1.5.
These Chandra X-ray Observatory detections of normal galaxies at high redshifts have initiated the study of the X-ray evolution of normal galaxies over cosmologically interesting distances. Evolution of the X-ray properties of galaxies is to be expected because the star formation rate (SFR) of the universe was at least a factor of 10 higher at redshifts of 1-3 (Madau et al. 1996). Since the X-ray luminosity of galaxies scales with the infra-red and optical luminosity (see Section 5; Fabbiano, Gioia & Trinchieri 1988; David, Jones and Forman 1992; Shapley, Fabbiano & Eskridge 2001; Fabbiano & Shapley 2002) the increased star formation will have a corresponding impact on the X-ray properties of galaxies at high redshift (White and Ghosh 1998). For spiral galaxies without an AGN, the overall X-ray luminosity in the 1-10 keV band will typically be dominated by the galaxy's X-ray binary population. There is expected to be a corresponding increase in the number of high mass X-ray binaries associated with the increased star formation rate. The `detection' of the Balmer and Lyman Break galaxies by Nanda et al. (2001) and the factor of 5 increase in the X-ray luminosity from redshift 1 to 3 is consistent with an increasing star formation rate. Nandra et al. (2001) point out that the X-ray luminosity of galaxies provides a new `dust free' method to estimate the star formation rate, as also pointed in the Beppo-Sax study of Ranalli, Comastri & Setti (2003), and by Grimm, Gilfanov & Sunyaev (2003).
The low mass X-ray binary (LMXB) population created by the burst in star formation at z > 1 may not emerge as bright X-ray sources until several billion years later (White and Ghosh 1998; Ghosh and White 2001). This is due to the fact that the evolutionary timescales of LMXBs, their progenitors, and their descendants are thought be significant fractions of the time-interval between the SFR peak and the present epoch. In addition to an enhancement near the peak (z 1.5) of the SFR due to the prompt turn-on of the relatively short-lived massive X-ray binaries, there may be a second enhancement, by up to a factor ~ 10, at a redshift between ~ 0.5 and ~ 1 due to the delayed turn-on of the LMXB population (Ghosh and White 2001). This second enhancement will not be associated with an overall increase in the optical or infrared luminosity of the galaxy, resulting in an increase in the X-ray to optical luminosity ratio. Hornschemeier et al. (2001) using the `stacking' technique detected X-ray emission from L* redshift 0.4 to 1.5 spiral galaxies in the HDF-N. The X-ray to optical luminosity ratios are consistent with those of galaxies in the local universe (e.g., Shapley, Fabbiano & Eskridge 2001), although the data indicate a possible increase in this ratio by a factor of 2-3.
Ptak et al. (2001) discuss the observable consequences of the increased SFR at high redshifts for the X-ray detection of galaxies at redshift > 1 in the HDF-N. To do this Ptak et al. (2001) used the Ghosh and White (2001) models for the evolution of the underlying X-ray binary populations for several different possible SFR models (the SFR with redshift is not well known). Depending on the SFR model used, the average X-ray luminosity of galaxies in the HDF-N can be an order of magnitude higher than in the local universe. These model predictions can be translated into a prediction of the number counts verses flux. Fig. 14 taken from Hornschemeier et al. (2003) shows the number counts from the CDF-N (which are dominated by AGN), along with the predictions from Ptak et al. (2001) for two different SFR models. The emerging population of optically bright, X-ray faint (OBXF) galaxies detected in the CDF-N is also shown, along with the extension of the source counts to fainter fluxes using a fluctuation analysis of the CDF-N (Miyaji & Griffiths 2002). The predictions are that emission from normal galaxies, largely at redshift of 1-3, will start to dominate the source counts somewhere between fluxes of 10-17 and 10-18 erg cm-2 s-1 . The cross on Fig. 14 shows the constraint from the stacking analysis of Hornschemeier et al. (2002) for relatively nearby spiral galaxies (z < 1.5), which is in agreement with the predictions from Ptak et al. (2001) for the lower SFR models.
Figure 14. CDF-N number counts, with predictions (ar the faint end) based on different SRF at high redshift. The open boxes are the counts from the optically bright, X-ray faint sources - these are mainly normal and starburst galaxies, but some low-luminosity AGN may be present -. The cross is the result of the `stacking' analysis using z 1.4 galaxies in the CDF-N field. The solid and dashed lines at the lowest fluxes are the the predictions of the galaxy number counts from Ptak et al 2001 (from Hornschemeier et al 2003). The leftward pointing arrow indicates the number density of field galaxies at I = 24 mag.
Much deeper Chandra exposures of several months or even a year long will be able to eventually reach fluxes of 10-18 erg cm-2 s-1 and directly test the models for the X-ray evolution of galaxies - given the projected long lifetime of Chandra and good luck, these very deep exposures will hopefully will eventually happen as the mission matures. To obtain spectra of these galaxies, which are typically at a redshift of 1-3, and to see higher redshift objects at a similar faint flux level, will require 100-1000 times more collecting area with 1 arc sec angular resolution to avoid confusion (e.g., Fabbiano 1995, 2000; Elvis & Fabbiano 1997; Fabbiano & Kessler 2001). Even more challenging, to resolve an AGN or an offset ULX from more extended emission from the galaxy will require an angular resolution of order 0.1 arc sec. Such mission parameters are technologically extremely challenging, but nonetheless are being pursued by NASA, ESA and ISAS as a long term goal for X-ray astronomy (Parmar et al. 2002, Zhang et al. 2001).