«CRAIG J. HOGAN Departments of Physics and Astronomy, University of Washington, Box 351580, Seattle, WA 98195, USA Abstract. Estimates of the ...»
EXTRAGALACTIC ABUNDANCES OF HYDROGEN,
DEUTERIUM AND HELIUM
New Steps, Missteps and Next Steps
CRAIG J. HOGAN
Departments of Physics and Astronomy, University of Washington, Box 351580, Seattle,
WA 98195, USA
Abstract. Estimates of the deuterium abundance in quasar absorbers are reviewed,
including a brief account of incorrect claims published by the author and a brief review
of the problem of hydrogen contamination. It is concluded that the primordial abundance may be universal with a value (D/H)P ≈ 10−4, within about a factor of two, corresponding to ΩB h2 ≈ 0.02 or η10 ≈ 2.7 in the Standard Big Bang. This agrees with current 0.7 limits on primordial helium, YP ≤ 0.243, which are shown to be surprisingly insensitive to models of stellar enrichment. It also agrees with a tabulated sum of the total density of baryons in observed components. Much lower primordial deuterium (≈ 2 × 10−5 ) is also possible but disagrees with currently estimated helium abundances; the larger baryon density in this case ﬁts better with current models of the Lyman-α forest but requires the bulk of the baryons to be in some currently uncounted form.
1. Introduction Material in the Earth and other planets, the solar neighborhood, and indeed our entire Galaxy has experienced signiﬁcant chemical evolution that has modiﬁed the traces of light elements from the Big Bang. We are now beginning to sample abundances in more distant environments with a variety of diﬀerent chemical histories. Some are nearly pristine and serve as fossil beds preserving the original chemistry, particularly those at at high redshift before much of the primordial gas ﬁrst formed into galaxies and stars. In addition to insights about chemical evolution under diﬀerent circumstances the new measurements give us unique information about the history of ﬁne-grained structure over an enormous spacetime volume; for example they test the idea that primordial chemistry is the same everywhere and that the primordial gas was precisely uniform on small scales. New techniques also now allow us to tabulate a more reliable direct estimate of the mean baryon density of the Universe in various forms, allowing a further test of Standard Big Bang Nucleosynthesis for which mean total baryon density is the principal parameter.
The theory and its concordance have been extensively reviewed in the literature from many points of view (e.g. Walker et al. 1991, Smith et al. 1993, Copi et al. 1995, Sarkar 1996, Fields et al. 1996,, Hogan 1997, Schramm 1998), and many of the topics covered here are reviewed in more detail 2 C. HOGAN elsewhere in this volume. This review explores a selection of interesting contradictions among new extragalactic datasets but also highlights the general concordance with the Standard Big Bang Model.
The measurement of deuterium abundances by analyzing Lyman series absorption lines in stellar spectra from foreground diﬀuse gas (Rogerson and York
1973) has yielded a reliable abundance for the Galactic interstellar medium, accurate to about ±50%. Thanks to new technology the same technique applied to quasar spectra now yields the abundance of much more distant gas (see Tytler 1998, Vidal-Madjar 1998). High resolution, high signal-tonoise spectra allow the study of resolved absorption features of low optical depth which accurately count atomic column densities and thus in principle give reliable abundances. The subject is still very young however; eﬀects of ﬁnite resolution and saturation still contribute a subtle source of error (see Levshakov et al. 1998), and the systematic errors are not yet well calibrated since the physical model of the absorbing material is not mature. (It is certainly not, as assumed in the spectral ﬁts, in discrete isothermal slabs).
At the moment there is an apparent polarization between “high” and “low” values which seems likely to disappear with time.
2.1. Evidence for a High Abundance
An early Keck spectrum of QSO 0014+813 allowed the ﬁrst estimate of an extragalactic deuterium abundance, yielding a remarkably high value D/H ≈ 2 × 10−4 (Songaila et al. 1994). Multiple lines of the Lyman series gave a good agreement of hydrogen and deuterium redshift, and a reliable estimate of both column densities, so the most signiﬁcant uncertainty in this measurement was the amount of contamination of the deuterium absorption feature by interloping hydrogen.
Subsequent analysis of this same data (Rugers and Hogan 1996a) showed that at full resolution the Lyman-α feature of deuterium was split into two components, each of which was too narrow for hydrogen. The authors used this fact to argue that the hydrogen contamination did not dominate the deuterium feature and therefore that the high abundance could be trusted.
The split of the feature was traced back to an electron excess in the raw echellogram data so was not a reduction artifact.
However, in a classic illustration of systematic error it now appears that the feature was not a real spectral feature of the quasar. Subsequent spectra by both the Hawaii group and by Tytler et al. (1997), which have better signal-to-noise than the original spectrum, do not conﬁrm it. Apparently
9712031.tex; 14/01/1998; 2:59; no v.; p.2 EXTRAGALACTIC ABUNDANCES
the ≈ 60 electrons contributing to the earlier signal, whatever their origin, were neither simply noise nor photoelectrons from the quasar light. (Similar problems also appear to plague another candidate feature in the same spectrum, proposed by Rugers and Hogan 1996b). The deuterium feature in the real spectrum is instead smooth and is well ﬁt by a single thermally broadened component, so the linewidth is consistent with signiﬁcant hydrogen contamination. On the other hand the width is the same as the associated hydrogen (24 km/sec) and is therefore also consistent with being caused by deuterium if the broadening is mostly turbulent.
Indeed, the better data allow the redshift of the deuterium and hydrogen features to be compared more precisely and they diﬀer by 10 km/sec, indicating that there is some hydrogen contaminating the deuterium (Tytler et al. 1997). However the total column density of contaminating hydrogen required to move the centroid is small, so the known presence of some contamination does not imply that the best estimate of the abundance changes appreciably from that originally given by Songaila et al.
Because of the good constraints on both hydrogen and deuterium column densities, and in spite of confusing claims made by this author, the new data on the Q0014+813 absorber still displays good evidence of a high primordial abundance, about as convincing as the original claim by Songaila et al.
(1994). In another quasar (BR 1202-0725) a “detection or upper limit” at D/H = 1.5 × 10−4 was found by Wampler et al. (1996). A high abundance (2 × 10−4 ) may also be detected in Q0402-388, although it is required only if OI/HI is assumed to be constant in the ﬁtted components (Carswell et al. 1996). The same high abundance (2 × 10−4 ) was also found by Webb et al. (1997) in Q1718+4807, although this is also not yet conclusive; the present analysis relies on a SiIII line to ﬁx the redshift of the hydrogen, the D column is based only on a Lyman-α ﬁt and the H column on a low resolution spectrum of the Lyman limit. The agreement between these estimates is at least suggestive of a high universal abundance, although none of the evidence is yet conclusive.
2.2. Evidence for a Low Abundance
Burles and Tytler (1996) and Tytler et al. (1996) have presented evidence for a low D/H in two quasar absorbers. Of these the stronger case at present is in Q1937-1009 since high quality data are available up to the Lyman limit.
The estimated abundance is 2.3 ± 0.3 × 10−5, nearly an order of magnitude less than the high values discussed above.
Unfortunately the total column in this case is high so the HI absorption is optically thick even past the Lyman limit and the column density must be estimated from saturated features. This has led to a debate in the literature on the allowed range for the HI column and for D/H (e.g. Songaila
9712031.tex; 14/01/1998; 2:59; no v.; p.34 C. HOGAN
et al. 1997, Burles and Tytler 1997, Songaila 1997), which is likely to end up somewhere in the middle: although the total error in the abundance is probably larger than originally quoted by Tytler et al. (1996), the HI column is probably well enough constrained to exclude very high values like those in Q0014+813.
Therefore the dispersion in abundance between the best high and low estimates appears on the surface to be real. What is going on? There have been many suggestions (e.g., Jedamzik and Fuller 1996). Perhaps the primordial abundance is not uniform; perhaps the low-D systems have experienced stellar destruction of their deuterium; perhaps the high-D systems have found some exotic source of nonprimordial deuterium. The most prosaic explanation however is that the high-D features are all dominated by contaminating hydrogen.
2.3. Statistical Approaches to the Contamination Problem
The latter possibility deserves serious attention since the eﬀect is known to be there and known to bias abundance estimates upwards. However, a quantitative estimate of the eﬀect shows that it is unlikely to be important most of the time. The mean number of hydrogen lines in a velocity interval δv per ln(N [HI]) interval can be estimated from the line counts of Kim et al. (1997) to be about −0.4 δv N [HI] δP (N [HI]) ≈ 5 × 10−3.
10km sec−1 1013 cm−2 If the contaminating hydrogen lines are distributed at random, we can use Poisson statistics to estimate the probability of contamination. For example, in the case of Q0014+813 the column density of the DI feature is about
1013.2 cm−2 ; the probability of an HI line close to this column density appearing in the right redshift range to mimic deuterium (within an interval of about δv ≈ 20km sec−1 ) is only about P ≈ 1%.
Of course, smaller amounts of contamination are more likely. They tend to bias the deuterium abundance estimates upwards and create a nongaussian (power-law) error distribution allowing low D/H with nonnegligible probability. However the magnitude of the bias is still small in this range of column density; for example, the chance of a ≈ 10% contamination (for which there is indeed some evidence in Q0014+813 in the line proﬁles) is greater than the probability of 100% contamination by a factor of about 100.4 ; but this is still only a few percent.
Furthermore this calculation does not yet allow for the additional coincidence required in the Doppler parameters. Real deuterium cannot be wider than its corresponding hydrogen. Contaminating hydrogen has a linewidth drawn at random from the parent population, which is not compatible with
9712031.tex; 14/01/1998; 2:59; no v.; p.4 EXTRAGALACTIC ABUNDANCES
a deuterium identiﬁcation if it happens to exceed the width of the corresponding hydrogen feature. In the case of Q0014+813, since the D feature linewidth of 24 km sec−1 is typical of HI Lyman-α forest lines we should multiply the probabilities by about half to account for this eﬀect. This factor is smaller in situations where the features are unusually narrow.
It is important to verify that the lines are uncorrelated as we have assumed. Although there are disagreements over the amplitude of line correlations the best data on the Lyman-α forest shows correlations smaller than unity at velocity separations of 100km sec−1 ; for example in the line lists of Kim et al. (1997) the amplitude is less than 10% for all lines. We have checked this also speciﬁcally for correlations with high HI column density features similar to those studied in the D candidates; although the sample is smaller and the result noisier, the amplitude of the correlation is still less than a few tenths. In this study we used one-sided correlations (counting line companions only to the red side of each Lyman-α line) in order to exclude ﬁrst-order eﬀects of deuterium contamination in the hydrogen sample. (Similar conclusions were recently reached by Songaila 1997).
Thus even without a thorough physical understanding of the absorbing material, these empirical studies indicate that correlations among the lines are too weak to alter signiﬁcantly the simple estimates made on the basis of Poisson statistics. Therefore the best guess is that the deuterium abundance in cases where it appears to be high really is high, at least where the DI column is greater than about 1013 cm−2 (which is in any case required for a reasonably precise column estimate at realistic signal-to-noise). Noting that the full error in the ﬁtting technique has not yet been calibrated on realistic physical models of the clouds, it still seems reasonable to guess that the current data are consistent with a universal primordial abundance with a factor of two of 10−4. A much more convincing measurement will be possible with a few more good targets.
2.4. Next Steps
The contamination problem is less for lower redshift, where the Lyman-α forest thins out. Although low redshift quasar spectroscopy at high resolution is costly as it requires a large investment of time with the Hubble Space Telescope, the new STIS two-dimensional spectrograph can observe the entire Lyman series at the same time; the greater eﬃciency will give more solid results on cases such as Q1718+4807 which are already known to be interesting (e.g. Webb et al. 1997).
Contamination can also be reduced by studying high column density systems. Especially interesting are damped HI absorbers where reliable columns can be obtained for both HI and DI (Jenkins 1996), although targets are also rarer and tend to be in evolved galaxies. The best target of this type
9712031.tex; 14/01/1998; 2:59; no v.; p.56 C. HOGAN
so far identiﬁed is Q2206-199, which has a low metal abundance and a very low velocity dispersion (Pettini and Hunstead 1990).