Are the observed Ly forest clouds the main source of the observed strong He II opacity? This is addressed in Fig. 4 where the normalized high-resolution optical Ly forest spectrum (scaled in wavelength according to 303.78/1215.67) is overlayed the normalized He II "forest" spectrum.
We note three different types of He II absorption regions:
We have used the column densities, b -values and redshifts of the detected H I clouds to predict the He II absorption, adopting either turbulent or thermal line broadening. Simulated spectra were degraded to the GHRS resolution for a comparison with the observation. It is clear already from Fig. 4 that the three components ("voids", "troughs" and the associated system) cannot be modelled by scaling the Ly forest clouds with one constant column density ratio .
This is shown in Fig. 5 where the model calculation assumes = 100 and , i.e. pure turbulent broadening, which produces maximum He II opacity. While the two "voids" around 1160 and 1174 Å are perfectly matched - notice the structure within the 1160 Å void and the strong line on the short wavelength side of the 1174 void - the troughs cannot be modelled with the observed Ly forest clouds, even for = 1000. A continuous H I component with = 0.01 or a quasicontinuous blend of weak lines like the "undulating" absorption observed by Tytler (1995) in high S/N spectra of HS 1946+7658 could yield the observed He II opacity for . However, whatever combinations of and broadening parameters b are chosen, the "voids" and "troughs" cannot be modelled with the same parameter set.
As was shown already by Hogan et al. (1997), the observed Ly clouds in the high-resolution optical spectrum of Q 0302-003 alone cannot explain the absorption even for extremely soft ionizing spectra. Since our optical data do not allow the detection of hydrogen clouds down to log cm-2, we performed Monte Carlo simulations to estimate the amount of He II absorption by the weakest Ly forest clouds. The distribution of Ly clouds in redshift and column density was described by
As far as the associated system is concerned, we expect a much harder ionizing spectrum due to the quasar, i.e. lower values of . Considering both heavy element and hydrogen absorption lines in the optical data we can distinguish 14 individual components of the associated system ranging from to 2.9047, i.e. spanning km s-1. For absorber components with the strongest H I absorption (i.e. those at ) the observed He II absorption is best modelled adopting for , while at lower redshift ( -2.8985) is more appropriate. According to our estimates for the spectral shape of the intrinsic quasar energy distribution (see below) we would expect .
We conclude from the above that the ionization structure of the associated systems also cannot be modelled by a single set of parameters (cf. discussion in Sect. 3.4).
3.1.1. Pure Ly forest opacity
Fig. 5 shows that the "void" at 1160 Å can be explained by photoionized forest clouds far from the QSO with and turbulent broadening . This high value of is barely consistent with models of an ionizing metagalactic background due to QSOs which predict (Haardt & Madau 1996). However, the faint end of the QSO luminosity function is not known for and there may be a contribution by stars (Rauch et al. 1997). This "void" component is consistent with what Davidsen et al. (1996) found at lower redshift in HS 1700+6416.
3.1.2. Ly forest plus diffuse component
Assuming instead (Haardt & Madau 1996) and pure thermal broadening of He II, for which there is evidence from theoretical Ly forest simulations (e.g. Zhang et al. 1997), the known Ly forest clouds in the 1160 Å void contribute only 50 % of the observed opacity and an additional optical depth is required (see Fig. 6). The latter includes the contribution by faint optically thin forest lines not detected in our high-resolution spectrum. According to Madau & Meiksin (1994) the optical depth of the diffuse He II gas is related to the background intensity in photoionization equilibrium by
where is the mean intensity of the ionizing background at 912 Å in units of 10-21 erg cm-2 s-1 Å-1 sterad-1, h50 is the Hubble constant in units of H0 = 50 km s-1 Mpc-1 and q =0.5. With , , ( /1.8) Eq. (2) yields h . Because of the assumed lower limit to a He II forest contribution, this must be considered as a strict upper limit.
Our result for the void at 1160 Å is also consistent with the diffuse gas density derived by Hogan et al. (1997) for the void in Q 0302-003 at . They found 2 1.3 for the diffuse component. With the same parameters used in this paper this translates to 0.043 h 0.035 h for = 45 and 0.029 h 0.023 h for = 100, respectively. The latter softer ionizing spectrum appears more appropriate at (Songaila & Cowie 1996) and is consistent with the diffuse density 0.02 h derived at in HE 2347-4342. With the same shape of the ionizing background at and , the diffuse density found in the HE 2347-4342 void is lower by a factor of 2 than in the Q 0302-003 void.
If the voids are caused by additional ionizing sources - and we have no better explanation for voids in the Ly forest - the background radiation may be irrelevant for the ionization in the voids. In case of an AGN as ionizing source, would be smaller () and the necessary amount of diffuse gas would increase. A star dominated ionizing source, on the other hand would mean a larger () and no diffuse medium would be required. We must conclude that at present we have no real constraints for a diffuse medium in the voids.
3.2. The "troughs"
The mean flux measured from our data in the wavelength range from 1175.5 to 1181 Å is erg s-1 cm-2 Å-1. The continuum flux measured just longward of the He II edge is erg s-1 cm-2 Å-1. The corresponding flux ratio implies a high optical depth . When the flux depression by discrete He II absorption of Ly forest clouds is taken into account the minimum required optical depth is to explain the vanishing flux.
If the same ionizing UV background is responsible for the ionization in the "troughs" the density of the diffuse gas can easily be estimated. According to Eq. 2, yields h ( (0.023) h for ) close to the total big bang nucleosynthesis baryon density = 0.05 h (Walker et al. 1991). Assuming that the He II opacity in the "troughs" represents the normal IGM at and the "void" at 1160 Å is caused by higher ionization due to additional local ionizing sources, we are forced to conclude that the baryons are largely in the diffuse or at least the low density part of the IGM. This would be at variance with all recent hydrodynamical simulations of hierarchical structure formation (Cen et al. 1994; Miralda-Escudé et al. 1996; Zhang et al. 1997; Rauch et al. 1997; Meiksin 1997) which predict that except a few percent all of the baryons are in the Ly forest clouds, i.e. fragmentation of the IGM should be nearly complete.
There is a further argument against photoionization of the trough component by the same ionizing radiation as the voids. Since H I) He II)/ , the Gunn-Peterson constraints He II) (4.8) and H I) lead to (380), which is inconsistent with an AGN dominated background at for which there is ample evidence (e.g. Haardt & Madau 1996).
Trying to explain the observed flux depression, e.g. at 1180 Å, we also estimated the contribution from further absorber candidates like a) He II absorption from weak Ly clouds which are not detectable in our data, b) higher Lyman series lines from Ly clouds at low redshifts, c) neutral helium absorption or d) heavy element absorption lines. Of course none of these absorption lines alone can account for the total absorption observed.
In our calculation we considered at the same time a) a diffuse He II absorbing component with =0.2, b) He I 584 absorption from the strong Ly clouds detected at , c) He II absorption lines from presumed weak Ly forest clouds with log adopting in addition to He II absorption from detected Ly forest clouds. Unless one assumes all these absorbers cannot explain the vanishing flux.
Further absorption turning on just at the He II edge is expected by the strong O III doublet at 303 and 305 Å. Down to it is possible to identify MLS by C IV absorption longward of Ly emission in the optical data. Besides the associated system we identified 9 MLS, which often split into several components. For absorbers with the O III 303,305 doublet falls longward of 1150 Å. We calculated simple CLOUDY (Ferland 1993) models to estimate their influence. Unless adopting very unrealistic high column densities for O III these lines cannot explain a total absorption.
3.3. Delayed He II ionization
We see only one way to resolve the observed inconsistencies: He II ionization in the "trough" component has to be incomplete. Meiksin & Madau (1993) have predicted H I absorption "troughs" with vanishing flux, velocity broadened by the Hubble expansion. This has so far not been seen in H I up to . We propose here that we observe a patchy intergalactic medium in which the reionization of the universe is still incomplete at 3 for the second ionization of He. This possibility has been predicted by various authors, e.g. Miralda-Escudé & Rees (1993) and Madau & Meiksin (1994). The reason for the delayed ionization of He II is the much smaller number of He II ionizing photons compared to H ionizing photons. Madau & Meiksin (1994) show that with QSO spectra varying as and the quasars turned on at , hydrogen is fully ionized at , but that He II is only partially ionized at . According to recent work by Shaver et al. (1996), QSOs are an order of magnitude less abundant at than at , i.e. the turnon of the QSOs as ionizing sources is even later than assumed by Madau & Meiksin (1994). Furthermore, there is evidence from modelling of the Ly forest that for the UV-background due to QSOs is not sufficient and that additional sources of photons are required for (Rauch et al. 1997). If these sources are stars, the resulting softer UV background possibly meets the condition for delayed He II ionization. In addition, it has been shown by Giroux & Shapiro (1996) that in case of an IGM partially collisionally ionized due to bulk heating in addition to photoionization, He ionization will be lag behind hydrogen ionization. There is also evidence of an abrupt change in the C IV /Si IV column density ratio around (Songaila & Cowie 1996), which indicates a softening of the ionizing UV background with increasing z: to 40 below and above .
In case of only singly ionized helium, the amount of diffuse gas needed to produce the "trough" opacity is given by
for q (Madau & Meiksin 1994) which leads to 1.3 10-4 h . Only 0.3 % of the big-bang nucleosynthesis baryon density = 0.05 h is required to produce the blacked-out troughs. This would be consistent with the theoretical predictions that fragmentation of the IGM is nearly complete.
It is easy to show that even for a soft ionizing spectrum ( = 2) the "troughs" will not be seen in He I 584 Å since for = 0.02 (cf. Jakobsen 1995). The FOS spectrum around 2240 Å confirms the absence of a He I trough.
3.4. No proximity effect?
An extremely luminous QSO like HE 2347-4342 is expected to show a proximity effect except for the rather improbable case that the QSO has been turned on only recently.
The influence of the UV radiation of the QSO itself on the second ionization of He has been studied theoretically by various authors (e.g. Zheng & Davidsen 1995; Meiksin 1995; Giroux et al. 1995). The prediction is that close to the QSO He II is additionally ionized and that the effect is stronger in diffuse than in clumped gas. Consequently, if the He II opacity close to the QSO is dominated by a diffuse component, we expect increased transparency in a He III bubble around the QSO, close to the rest redshift of the QSO, while the effect will be weaker in the Ly forest. No proximity effect has been seen in HS 1700+6416 (Davidsen et al. 1996), while Hogan et al. (1997) observed in Q 0302-003 a "shelf" in the He II opacity 12 Å blueward of the main He II edge which they interpret as a proximity effect in diffuse He II gas. If, as concluded above, the "troughs" in the He II forest at 1162-1173 Å and 1176-1182 Å are caused by not fully ionized gas, we should expect that HE 2347-4342, which at the time of observation is one of the most luminous sources in the universe, produces a strong proximity effect.
In the following we estimate the He II ionizing flux of HE 2347-4342 and discuss its effect on the surrounding medium. UV spectra of high-redshift quasars are strongly influenced by absorption of intervening absorbers. In order to find the intrinsic spectral energy distribution of the quasar corrections have to be applied to the observed data. In the following we will consider only the flux depression by cumulative hydrogen continuum absorption of intervening absorbers, neglecting effects of dust absorption along the line of sight or line blanketing in the UV. Uncertainties in the absolute flux calibrations of the spectra up to 10-20 % are possible.
The dereddened spectra of HE 2347-4342 were corrected for continuum absorption by neutral hydrogen in the strong LLS at . The continuously rising spectrum of HE 2347-4342 shows no evidence for further strong absorbers with log . The additional "Lyman valley" depression of the quasar continuum due to the cumulative hydrogen absorption by the numerous Ly clouds with log was estimated by Monte Carlo simulations. Since Å is not directly observable, we take the corrected continuum value at the smallest observed wavelength ( Å), assuming that it is close to the flux at 228 Å. We find (885.78 Å) (1157 Å) = 6.4 10-15 erg s-1 cm-2 Å-1, i.e. nearly a factor of 2 higher than the observed flux.
For the QSO flux at Å we considered the flux in the range from 3560 to 3580 Å in the optical data just longward of the corresponding but unobserved wavelength Å. Since in low-resolution optical data the flux is depressed by unresolved line blanketing of Ly forest lines, we rebinned our high-resolution optical data to estimate the flux depression by line blending. The resulting flux is then Å)=1.75 10-15 erg s-1 cm-2 Å-1 yielding a ratio of . Adopting a pure power law () for the intrinsic EUV spectral energy distribution turns into , which is a typical value for several luminous QSOs observed with HST (Köhler & Reimers 1996). In the proximity of the quasar we thereby do not expect a strong contribution from diffuse Helium. With
(Madau & Meiksin 1994) and (Giallongo et al. 1994) we expect for an optical depth of the diffuse component comparable to derived from the observations. For the associated system we would expect () also consistent with the observation at least for the components at highest redshifts when adopting turbulent line broadening.
For a QSO flux at the He II edge of erg s-1 cm-2 Hz-1 at (for Å)=6.4 10-15 erg s-1 cm-2 Å-1 as derived above and Eq. 6 from Giroux et al. 1995) we get erg s-1 cm-2 Hz-1 at (corresponding to He II 304 absorption at 1160 Å) which has to be compared with a background flux erg s-1 cm-2 Hz-1 (Haardt & Madau 1996) or erg s-1 cm-2 Hz-1 (Bechtold 1995 with ). This means that HE 2347-4342 with its observed hard () EUV spectrum dominates He II ionization at least up to the Å void. This is not observed. Why does HE 2347-4342 ionize He II in the associated system but not in the troughs, in particular the region -1180 Å? There is only one plausible reason for the nonexistence of a proximity effect in He II: shielding by clouds optically thick in the He II 228 Å continuum along the line of sight. Possible candidates are the associated system clouds.
A reliable estimate of the neutral hydrogen column densities of the individual components of the associated system is difficult due to the complex structure of the system and saturation effects. Higher Lyman series lines up to H I 920 are visible, but the signal to noise ratio of our spectrum decreases rapidly at shorter wavelengths. According to absorption line fits the largest column density for a single component should not exceed log for km s-1. From optical data we cannot preclude the existence of a small break at the Lyman edge expected at 3550 Å. The maximum total H I column density compatible with the optical data is then cm-2. With the H I column densities derived from modelling of Lyman series lines adopting km s-1 for the 14 individual components the observed He II absorption requires for thermal line broadening. Thus for a single component with log and we find cm-2. The clouds become optically thick for He II at a column density 6.3 1017 cm-2 and self shielding could enhance the He II absorption. In fact, as was mentioned in Sect. 3, gives a better fit for the lower redshift components ( -2.8985) which leads to possibly cm-2, sufficient to shield the IGM in our direction from He II ionizing photons. A detailed analysis of the associated system with better optical data will help to improve the constraints for the He II continuum absorption.
What is the origin of the 1160 Å and 1174 Å voids? If the troughs are due to unionized He II which completely shields the EUV background at Å, we have to assume that individual sources (QSO and/or starburst galaxies) create He III regions within the unionized IGM. With a diameter of 2.8 h Mpc for the 1160 Å void, the requirements on the luminosities of possible sources are moderate (cf. Bajtlik et al. 1988, Giroux & Shapiro 1996). The observed would favour at least a contribution by stars in the ionizing source.
© European Southern Observatory (ESO) 1997
Online publication: April 6, 1998