The abundance of deuterium (D or 2H) is the most sensitive measure of the baryon density . No known processes make significant D, because it is so fragile (, , , and ). Gas ejected by stars should contain zero D, but substantial H, thus D/H decreases over time as more stars evolve and die.
We can measure the primordial abundance in quasar spectra. The measurement is direct and accurate, and with one exception, simple. The complication is that the absorption by D is often contaminated or completely obscured by the absorption from H, and even in the rare cases when contamination is small, superb spectra are required to distinguish D from H.
Prior to the first detection of D in quasar spectra , D/H was measured in the ISM and the solar system.Ā† The primordial abundance is larger, because D has been destroyed in stars. Though generally considered a factor of a few, some papers considered a factor of ten destruction . At that time, most measurements of 4He gave low abundances, which predict a high primordial D/H, which would need to be depleted by a large factor to reach ISM values .
Reeves, Audouze, Fowler & Schramm  noted that the measurement of primordial D/H could provide an excellent estimate of the cosmological baryon density, and they used the ISM 3He +D to concluded, with great caution, that primordial D/H was plausibly 7 ± 3 x 10-5.
Adams  suggested that it might be possible to measure primordial D/H towards low metallicity absorption line systems in the spectra of high redshift quasars.Ā† This gas is in the outer regions of galaxies or in the IGM, and it is not connected to the quasars. The importance of such measurements was well known in the field since late 1970s , but the task proved too difficult for 4-m class telescopes (, , ) The high SNR QSO spectra obtained with the HIRES echelle spectrograph  on the W.M. Keck 10-m telescope provided the breakthrough.
There are now three known absorption systems in which D/H is low: first, D/H = 3.24 ± 0.3 x 10-5 in the zabs = 3.572 Lyman limit absorption system (LLS) towards quasar 1937-1009 , ; second, D/H = 4.0+0.8-0.6 x 10-5 in the zabs = 2.504 LLS towards quasar 1009+2956 , and third, D/H < 6.7 x 10-5 towards quasar 0130-4021 . This last case is the simplest found yet, and seems especially secure because the entire Lyman series is well fit by a single velocity component.Ā† The velocity of this component and its column density are well determined because many of its Lyman lines are unsaturated. Its Ly line is simple and symmetric, and can be fit using the H parameters determined by the other Lyman series lines, with no additional adjustments for the Ly absorption line.Ā† There is barely enough absorption at the expected position of D to allow low values of D/H, and there appears to be no possibility of high D/H. Indeed, the spectra of all three QSOs are inconsistent with high D/H.
There remains uncertainty over a case at zabs = 0.701 towards quasar 1718+4807, because we lack spectra of the Lyman series lines which are needed to determine the velocity distribution of the Hydrogen, and these spectra are of unusually low signal to noise, with about 200 times fewer photons per kms-1 than those from Keck. Webb et al. ,  assumed a single hydrogen component and found D/H = 25 ± 5 x 10-5, the best case for high D/H. Levshakov et al.  allow for non-Gaussian velocities and find D/H ~ 4.4 x 10-5, while Tytler et al.  find 8 x 10-5 < D/H < 57 x 10-5 (95%) for a single Gaussian component, or D/H as low as zero if there are two hydrogen components, which is not unlikely. This quasar is then also consistent with low D/H.
Recently Molaro et al.  claimed that D/H might be low in an absorber at z = 3.514 towards quasar APM 08279+5255, though they noted that higher D/H was also possible. Only one H I line, Ly, was used to estimate the hydrogen column density NHI and we know that in such cases the column density can be highly uncertain. Their Figure 1 (panels a and b) shows that there is a tiny difference between D/H = 1.5 x 10-5 and 21 x 10-5, and it is clear that much lower D is also acceptable because there can be H additional contamination in the D region of the spectrum. Levshakov et al.  show that NHI = 15.7 (too low to show D) gives an excellent fit to these spectra, and they argue that this is a more realistic result because the metal abundances and temperatures are then normal, rather than being anomalously low with the high NHI preferred by Molaro et al.
The first to publish a D/H estimate using high signal to noise spectra from the Keck telescope with the HIRES spectrograph were Songaila et al. , who reported an upper limit of D/H < 25 x 10-5 in the zabs = 3.32 Lyman limit system (LLS) towards quasar 0014+813.Ā† Using different spectra, Carswell et al.  reported < 60 x 10-5 in the same object, and they found no reason to think that the deuterium abundance might be as high as their limit.Ā†Improved spectra  support the early conclusions: D/H < 35 x 10-5 for this quasar. High D/H is allowed, but is highly unlikely because the absorption near D is at the wrong velocity, by 17 ± 2 km s-1, it is too wide, and it does not have the expected distribution of absorption in velocity, which is given by the H absorption. Instead this absorption is readily explained entirely by H (D/H 0) at a different redshift.
Very few LLS have a velocity structure simple enough to show deuterium. Absorption by H usually absorbs most of the quasar flux near where the D line is expected, and hence we obtain no information of the column density of D. In these extremely common cases, very high D/H is allowed, but only because we have essentially no information.
All quasar spectra are consistent with low primordial D/H ratio, D/H ~ 3.4 x 10-5. Two quasars (1937-1009 & 1009+2956) are inconsistent with D/H 5 x 10-5, and the third (0130-4021)Ā† is inconsistent with D/H 6.7 x 10-5. Hence D/H is low in these three places. Several quasars allow high D/H, but in all cases this can be explained by contamination by H, which we discuss more below, because this is the key topic of controversy.