BROAD H i ABSORBERS AS METALLICITY-INDEPENDENT TRACERS OF THE WARM-HOT INTERGALACTIC MEDIUM^*

Using thermally broadened H i lines to trace WHIM gas avoids many of the biases that can plague metal-ion-based WHIM tracers (see Danforth 2009). However, BLA surveys are fraught with observational complications as well (e.g., Richter et al. 2004, 2006). First, BLA lines are hard to detect and require spectra with a high S/N and minimal instrumental systematics. For example, at T = 10⁵ K, the hydrogen neutral fraction is approximately $f_{\rm H\,\mathsc {i}}\sim 10^{-5}$ (Figure 1) and the thermal b-value is 40 km s⁻¹ (Equation (1)). An IGM absorber with a total hydrogen (H i+H ii) column density N_H = 10¹⁸ cm⁻², typical of what is expected in the Lyα forest, would still have an observable H i column of $N_{\rm H\,\mathsc {i}}\sim 10^{13}$ cm⁻². In the optically thin regime (linear COG of growth), the equivalent width W_λ and line-center optical depth τ₀(N, T) for a Gaussian profile BLA are

$$\begin{eqnarray} W_\lambda (N)&=&\bigg (\frac{\pi e^2}{m_e c}\bigg)\,\bigg(\frac{N f \lambda ^2}{c}\bigg)=(54.5\;{\rm m\AA})\,N_{13}, \end{eqnarray} \tag{ 2 }$$

$$\begin{eqnarray} \tau _0(N,T)&=&\frac{\pi e^2}{m_e c} \frac{N f \lambda ^2}{\sqrt{\pi } b}=(0.187)\,N_{13}\,T_5^{-1/2}, \end{eqnarray} \tag{ 3 }$$

where N₁₃ is the H i column density in units of 10¹³ cm⁻². Thus, an IGM absorber with N₁₃ ≈ 1 will have a rest-frame Lyα equivalent width W_λ = 55 mÅ and a fractional depth of ∼ 20% (assuming no non-thermal broadening). This line would be easily observable in data of modest S/N.

Zoom In Zoom Out Reset image size

However, the prospects get much worse at higher temperatures, as the thermal line width rises as $\sqrt{T}$ and the neutral fraction drops (Figure 1). A Lyα absorber with the same total hydrogen column as above but with a temperature of 3 × 10⁵ K (near the peak CIE abundance of O vi) would have $f_{\rm H\,\mathsc {i}}\sim 10^{-6}$, $N_{\rm H\,\mathsc {i}}\sim 10^{12}$ cm⁻², b = 70 km s⁻¹, W_λ ∼ 5 mÅ, and a fractional depth τ₀ ∼ 1%, impossible to detect unless the spectra have S/N ≫100. Turbulent and/or bulk cloud motions would broaden the Lyα line even further and reduce the fractional line depth. Confirmation of a Lyα absorber and definitive measurements of $b_{\rm H\,\mathsc {i}}$ and $N_{\rm H\,\mathsc {i}}$ using higher Lyman lines and COG techniques are difficult for such lines since even Lyβ is a factor ∼ 7 weaker than Lyα for unsaturated systems.

Detecting low-contrast absorption features depends critically on obtaining a correct model of the continuum level, as well as a presumption that the intrinsic continuum emanating from the AGN central engine is flat and featureless over these broad line widths. Since the UV continuum of Seyferts and QSOs is thought to arise in thermal accretion disk emission close to the central black hole, this requires that the accretion disk itself be featureless, which may or may not be a correct assumption. For this reason, the UV continua of BL Lac objects may provide better background sources for sensitive BLA detection. Additionally, fixed-pattern noise and echelle-blazed gratings can themselves produce gentle undulations in the continuum, either on the detector or in the extraction process of curved orders imaged on rectilinear detectors and their corresponding artifacts and blemishes. The possible inability to place an accurate continuum due to the above difficulties not only makes low-contrast line detection suspect, but also makes it hard to set sensible detection limits as a function of b-value.

Even assuming data with sufficiently high S/N and a well-defined continuum, the identification of BLAs relies crucially on deconvolving the thermal line width b_T from the total observed line width b_obs. A simple, single-component absorber can be modeled as a Voigt profile, and the observed line width is the quadrature sum of the thermal line width b_T, the instrumental point-spread function (usually approximated as a Gaussian b_inst), and non-thermal broadening term b_NT. The latter term encompasses bulk turbulence, multiple unresolved velocity components, or any other condition that will broaden a line profile. Instrumental broadening is generally well determined and can be subtracted from the total line width. However, the relative proportion of the thermal and non-thermal contributions is typically degenerate for absorption in a single species.

Some simulations (e.g., Richter et al. 2006) suggest that turbulence contributes on average ∼ 10% to the total line width in broad absorbers, while other simulations suggest that it may be as much as 50% (Cen & Fang 2006). Observational studies have shown that the line width measured from a single Lyα line is generally greater than that determined through a multi-Lyman-series COG (Shull et al. 2000). For example, the so-called Virgo Cluster absorbers in the high-S/N GHRS spectrum of 3C 273 show a pair of Lyα components with measured b = 41 km s⁻¹ and 34 km s⁻¹ for the 1015 km s⁻¹ and 1590 km s⁻¹ IGM absorbers, respectively (Weymann et al. 1995). However, subsequent FUSE observations (Sembach et al. 2001) determined b_COG = 30 km s⁻¹ and 16 km s⁻¹ using three and eight higher-order Lyman lines, respectively.

While these absorbers make good individual examples, statistical studies show that b-values determined from line-width measurements of single lines, hereafter denoted b_LW, systematically overpredict the true line width. Shull et al. (2000) found that b_LW/b_COG ∼ 2 using Lyα and Lyβ measurements of 12 absorbers. Danforth et al. (2006) confirmed the low-z result with a larger sample (∼ 100 absorbers). Similar conclusions were reached by Songaila (1997, 1998, 2001) for high-z Lyα forest lines. We refine this relationship further using the large catalog of DS08. Out of ∼ 650 H i absorbers in DS08 plus additions, we measured 164 in multiple Lyman lines, and a COG analysis was performed giving more accurate values of $b_{\rm H\,\mathsc {i}}$ and $N_{\rm H\,\mathsc {i}}$. Figure 2 shows the distribution of b_Lyα and b_COG for this sample. Note that there is no clear correlation of b_LW/b_COG with column density and only a weak trend with b_LW. The median overprediction ratio is b_LW/b_COG = 1.26^+0.49_−0.25 with a mean of 1.35. We will use this result extensively in our analysis for the BLAs in which only Lyα measurements are available.

Zoom In Zoom Out Reset image size

Another impediment to BLA identification and measurement is that Lyα absorption lines often show ambiguous component structure. Several narrow components in a close blend can artificially broaden a line. Sometimes this will manifest itself as an asymmetric or obviously non-Gaussian line profile, but this blending can be undetectable even at arbitrarily high S/N and good spectral resolution. For this reason, even COG-determined b-values can overpredict the thermal line width.

To estimate the degree of line-width overprediction from COG analysis, we have modeled two simple two-component systems where the line width and relative centroid separation are varied (see also Danforth et al. 2006). In case 1 (Figure 3(a)), two narrow components, each with thermal line width b_T = 15 km s⁻¹, are blended into a single absorption line in Lyα and higher Lyman lines. The equivalent widths of each blended Lyman line are fitted with a COG. For component separation Δv < b_T, the inferred b_COG overpredicts b_T by < 20%, but at Δv>b_T, the overprediction rises to nearly 50%. This is largely independent of the relative column densities of the two lines. Depending on the strength of the line, the resolution of the spectrograph, and the quality of the data, even separations of Δv>b_T are not obvious in the line profile. Thus, a pair of narrow lines can easily be mistaken for a single broader system, even using a COG.

Zoom In Zoom Out Reset image size

In case 2 (Figure 3(b)), we combine a narrow system (b_T,1 = 15 km s⁻¹) with a broad absorber (b_T,2 = 2 b_T,1) and vary the relative line strengths and component separations. For components of similar column density, b_COG is close to the mean of the two components, and separating the components by (1–2) b_T,1 increases the composite b_COG by ∼ 20%–50% as in case 1. When one component is considerably stronger than the other, it dominates the b_COG solution. It is reassuring that the total column density of the composite system is conserved, independent of component separation and relative strength (Jenkins 1986).

Appropriate to case 2, many observed absorbers classified as WHIM due to their O vi absorption are multiphase in nature (e.g., Danforth et al. 2006; Danforth & Shull 2008; Tripp et al. 2008). A cool, photoionized component (T < 10⁵ K) is present at nearly the same velocity as a warm-hot component. We now estimate the broad Lyα profile expected to correspond with observed WHIM gas. For a typical O vi absorber with $N_{\rm O\,\mathsc {vi}}\approx (10^{13.5}\,{\rm cm^{-2}})\,N^{\rm (O\,\mathsc {vi})}_{13.5}$ at the peak CIE O vi abundance temperature (T ≈ T_max = 10^5.45 K), the resulting BLA should have total hydrogen column density

$$\begin{eqnarray} N_{\rm hot}^{\rm (H)}&=&\frac{N_{\rm O\,\mathsc {vi}}}{\langle f_{\rm O\,\mathsc {vi}} \rangle (4.90\times 10^{-5})\,\bigl (\frac{Z}{0.1\,Z_\odot }\bigr)} \nonumber \\ &=&(2.93\times 10^{18}\,{\rm cm^{-2}})\,N_{13.5}^{\rm (O\,\mathsc {vi})}\,\bigg (\frac{Z}{0.1\,Z_\odot }\bigg)^{-1}, \end{eqnarray} \tag{ 4 }$$

and, from Equation (1), $b_{\rm H\,\mathsc {i}}=(68.2\;{\rm km\;s^{-1}})\,(T/T_{\rm max})^{1/2}$. Here, we adopt an oxygen abundance (O/H) = (4.90 × 10⁻⁵)(Z/0.1 Z_☉) scaled to 10% of the solar metallicity (Asplund et al. 2005). Given a typical neutral fraction $f_{\rm H\,\mathsc {i}}\approx 1.4\times 10^{-6}$ at T ≈ T_max, the resulting BLA will have $N_{\rm H\,\mathsc {i}}\sim 5\times 10^{12}$ cm⁻² and τ₀ ∼ 0.05, easily overwhelmed by the narrower, stronger H i absorption from the photoionized component (typically $N_{\rm H\,\mathsc {i}}\approx 10^{13}\hbox{--}10^{15}$ cm⁻², b = 20–30 km s⁻¹, τ₀ ≫ 1). The BLA signature will be apparent only in the line wings and will require exquisite data (S/N ⩾30–40) and a very good knowledge of the instrumental point-spread function to recover.

Multiphase systems are difficult to identify individually, but there is statistical evidence for broad-plus-narrow H i systems (e.g., Tripp et al. 2001, 2008; Danforth & Shull 2008). DS08 compared the $b_{\rm H\,\mathsc {i}}$ distribution of 83 H i systems with O vi detections with another 273 having clean O vi nondetections ($N_{\rm O\,\mathsc {vi}}<10^{13.2}$ cm⁻²). For the O vi detections, the median and standard deviations were $b_{\rm H\,\mathsc {i}}=31\pm 15$ km s⁻¹, while the O vi nondetections show $b_{\rm H\,\mathsc {i}}=26\pm 13$ km s⁻¹. This slight difference (at a low confidence level) suggests that weak BLA lines might be present in the O vi systems and broadening their overall H i profiles. However, it is doubtful whether a difference between the two populations would be apparent without the O vi detection "sign posts." There are several good, individual examples of this observational signature (e.g., Tripp et al. 2000, 2001; Stocke et al. 2005, 2006).

Observing multiple Lyman lines and determining $b_{\rm H\,\mathsc {i}}$ rigorously through a COG tends to produce more accurate line-width measurements, but even here, multiple components can broaden a line width. Because the overestimate of the true b_T is smaller using a COG than for Lyα alone, we will adopt b_COG as our best estimator for absorber temperature. Examining higher-order Lyman lines cannot be counted on to solve the multi-component problem since most BLA absorbers are too weak to allow Lyβ to be detected and measured in all components. Nor will metal ions, which show intrinsically narrower lines, always help, since many BLAs are expected in regions of little or no metal enrichment. Thus, we must adopt a statistical approach to BLA verification.

With the above points in mind, we set out to determine what fraction of the reported BLAs are legitimately tracing warm-hot gas at T ⩾ 10⁵ K. To accomplish this, we have independently extracted, reduced, and scrutinized the STIS/E140M spectra (Table 1) used by Lehner et al. (2007, L07 hereafter) to search for BLAs potentially arising in the WHIM. Complete details of our data reduction method are given in DS08. Briefly, STIS/E140M data were uniformly reduced using CalSTIS v2.19. Line-free continuum regions were defined interactively in 10 Å segments of the data and fitted with low-order Legendre polynomials. Continuum fit uncertainty was taken as the standard deviation of points about the mean in the defined continuum region. This uncertainty was added in quadrature with those from photon noise and line-fit uncertainties.

The L07 work is the most complete and comprehensive look at BLAs currently available. In order to determine the number and b-value distribution of BLA absorbers in the local universe, we have adopted the same definition as L07 and most other studies, requiring T>10⁵ K corresponding to a hydrogen thermal b>40 km s⁻¹. The relationship between COG-determined line width and line width measured from a single line shows a median ratio b_LW/b_COG = 1.26^+0.49_−0.25 in the large DS08 sample, and the correlation of that ratio with either $N_{\rm H\,\mathsc {i}}$ or b_LW is poor (Figure 2). Given this ratio, an absorber with b_LW ≳ 50 km s⁻¹ (i.e., 1.26 × 40 km s⁻¹) would be more likely than not to have a b_COG ⩾ 40 km s⁻¹ (see Figure 2). Assuming that b_COG = b_T (but see discussion in Section 2.1). We find that T ⩾ 10⁵ K for absorbers with line width b_LW ≳ 50 km s⁻¹, not the b_LW ⩾ 40 km s⁻¹ used by most previous works. This difference is central to our analysis. So as a guideline, we established the following categories and criteria for BLAs, using our analysis combined with that in the literature.

Category A. Probable BLA. There is no definitive test for BLAs, but we classify as "probable" any system with well-defined b_COG ⩾ 40 km s⁻¹ or b_LW>60 km s⁻¹ confirmed by independent measurement of L07 and ourselves and with no obvious sign of multiple component structure (such as an asymmetric profile) in Lyα or higher Lyman lines or metal ions (if available). Fifteen systems fall within this category. Note that we do not a priori use the presence or absence of O vi, N v, or other WHIM sign posts as a BLA determinant, although several O vi detections are present in this sample (see below).

Category B. Possible BLA. Less likely than category A, the "possible" category includes systems with 40 km s⁻¹ ⩽ b_LW ⩽ 60 km s⁻¹. This category also holds intermediate cases, where an absorber shows ambiguous component structure or a high degree of uncertainty as to line width and/or continuum fit, but a broad H i system component cannot be reasonably ruled out. Forty-eight H i absorbers fall in this category.

Category C. Non-BLA. Nearly half (56/119) of the H i lines with b ⩾ 40 km s⁻¹ published by L07 or DS08 are most likely not BLAs by our assessment for a variety of reasons: measured b_COG < 40 km s⁻¹ or b_LW < 40 km s⁻¹; probable alternate line identification; obvious narrow component structure; or simply not detected as absorption features in our reduction of the data (i.e., we do not confirm the extraction and/or analysis done by or referenced in L07, see Figure 4). The last group, encompassing nine purported absorbers, occurs primarily in the spectrum of HS 0624+6907 (Aracil et al. 2006a, 2006b). In the course of our independent analysis, we identified a few BLA candidates that were not listed in L07 or other independent literature sources. In most cases, these lines turn out to be something other than Lyα and thus fall into category C, as they are not confirmed by two independent sources.

Zoom In Zoom Out Reset image size

Because the seven individual sightlines were analyzed originally by different authors, somewhat differing analysis procedures were used and then adopted by L07. Each is briefly discussed below, with notes that bear on the BLA identification process.

HE 0226−4110. Originally analyzed by Lehner et al. (2006), the spectrum is shown in that paper. No COG measurements are reported; instead some b-values are obtained by simultaneous fitting of several Lyman-series lines (see discussion below). Longward of 1690 Å, the continuum in this spectrum appears to "ripple" in both our extracted spectrum (Figure 5) that of Lehner et al. This gives rise to four reported BLA candidates (one with b_LW ∼ 150 km s⁻¹) in only Δz = 0.007. The periodic nature of these features is suspicious, especially given that these features occur the highest spectral orders which cut across many detector rows before extraction. However, no similar features appear in the highest spectral order (1690 Å < λ < 1710 Å) for AGN where z_Lyα>z_AGN.

Zoom In Zoom Out Reset image size

HS 0624+6907. Originally analyzed by Aracil et al. (2006a, 2006b), no spectra were published. These authors do not mention obtaining COG b-values except for a Lyα absorption complex at z ∼ 0.06. Therefore, we assume that all b-values quoted by L07 for this spectrum are either Lyα line-width values or simultaneous line-width fits to two or more Lyman lines as described for HE 0226−4110 above. Several of the BLAs reported by L07 in this sight line are at or near the locations of detector artifacts; we see no believable absorption at these locations based on our own reduction of the data. The reported lines may be a result of smoothing the data over these bad pixels. Longward of 1580 Å in this spectrum our extracted data does not match even the presence of some of the absorption lines listed by L07 (Figure 4). These unconfirmed lines are listed in Table 2 as non-BLAs for completeness, but are not otherwise analyzed.

PG 1116+215. Originally analyzed by Sembach et al. (2004), the spectrum is shown in that paper. Sembach et al. (2004) use COG b-values where available, although none apply to the BLAs in this sight line.

PG 1259+593. This sight line was originally analyzed by Richter et al. (2004) and the spectrum is shown in that paper. Richter et al. list several COG b-values for BLAs and we add several more.

PKS 0405−123. Originally analyzed by Williger et al. (2006, W06) and reanalyzed by L07. Neither W06 nor L07 report COG measurements for this sightline. The spectrum is shown in Williger et al. Despite mentioning that the reanalysis was required due to the large number of suspect BLA candidates, L07 b-values from simultaneous line fits are similar to the Lyα-only line widths reported by Williger et al. (2006). We report several COG b-values for this sight line and, similar to many other absorption systems we generally find b_COG ⩽ b_LW (see below).

H 1821+643. For spectral analysis, L07 refer to K. R. Sembach et al. (2010, in preparation) which does not appear to have been published yet. There is no mention of COG measurements in the L07 description, so we have assumed that the reported values are b_LW.

PG 0953+415. L07 report these BLAs based upon an unpublished Tripp et al. analysis. We have assumed that all quoted b-values are based on Lyα line-width measurements since our own b_LW values are similar.

One of the procedural differences between L07 and the current work is that where more than one Lyman line is detected in absorption in these spectra, we have used a b-value determined by the COG method whereas L07 uses a simultaneous line fit of all available lines. In general, the b_COG values are significantly smaller than those obtained by the simultaneous line-fitting method. To evaluate this difference, we have used the 10 absorbers in the PKS 0405−123 sight line analyzed independently by ourselves, L07, and W06. For all 10 systems, the W06 (b_LW) measurements agree to within the errors with the L07 (simultaneous line fits) b-values. Half of these measurements also show excellent agreement with b_COG. However, the other five systems show b_COG values significantly less than values obtained by the simultaneous line-fit method (for details see Table 2) and these disagreements are for the largest b-values reported by L07 in this sightline. Three systems (z = 0.09659, 0.16121, and 0.32500) have COG b-values with tight error bars half or less of the L07 reported values; two other systems (z = 0.17876 and 0.25861) have similar differences between reported values but, with large b_COG error bars and so are not inconsistent with the large line widths reported by L07. Based upon these 10 systems only, we find significant differences between COG b-values and simultaneous line-fit b-values in roughly half the cases, all in the sense that the COG b-values are significantly less (35%–50% the amount). When higher S/N FUV spectra from COS become available, tests using these two techniques should be made to determine which method is the most reliable as a function of line width and S/N.

In measuring metal-ion absorption lines, DS08 employ a 4σ detection limit, with equivalent width

$$\begin{equation} W_{\rm obs}\ge \frac{({\rm SL}) (\lambda /R)}{({\rm S/N})_\lambda }, \end{equation} \tag{ 5 }$$

where the instrumental resolving power R = λ/Δλ and SL is the significance level in standard deviations. However, both DS08 and other studies locate Lyα lines interactively, which does not follow the strict W_min argument used for lines of metal ions. Determining the significance level of the Lyα lines reported by DS08 based a posteriori on the observed equivalent widths and data S/N, we find SL > 10–15 for the weakest detections. Weaker features are certainly visible in the data, but they can reasonably be explained as fixed-pattern noise or other instrumental features. Indeed, DS08 use SL ⩾ 10 when determining the redshift path length Δz_Lyα in their absorber frequencies and cosmological calculations.

Tripp et al. (2008) point out that Equation (5) is strictly valid only for unresolved features. Since BLAs are several times wider than the instrumental resolution element for both STIS and FUSE data, Equation (5) is even less accurate. However, we argue that, even if the SL is not rigorously correct, it is still relatively correct, and it gives us a basis for equal comparison between lines. Broader lines will be shallower for the same equivalent width and thus each pixel will show less contrast from the continuum at full spectral resolution. Smoothing over the number of pixels equal to the line width (something the human brain does naturally) to first order yields no change in sensitivity for lines of the same equivalent width but different b-values.

Previous BLA studies used the quantity log (N/b) as a detection criterion, reasoning that broader lines require a higher column density (and hence equivalent width) to reach the same line-center optical depth. In particular, Richter et al. (2006) used log (N/b)>11.3 (for N in cm⁻² and b in km s⁻¹) as their detection threshold. For a constant column density, log (N/b) changes rapidly for narrow lines, but much more slowly at b>40 km s⁻¹. For example, log (N/b)>11.3 corresponds to log  N_Lyα>12.9 at b = 40 km s⁻¹ but only 0.2 dex (60%) higher at b = 65 km s⁻¹. We argue that setting a detection threshold based purely on equivalent width, W_Lyα(S/N)/Δλ>4, is roughly equivalent to setting one in (N/b) for BLAs. It should be noted that ∼ 75% of the possible and probable BLAs in our sample show log (N/b) ⩾ 11.3, and all have log (N/b)>11.0.

Total absorption path length Δz depends on the redshift of the background AGN and the wavelength coverage of the UV spectrograph. We follow a procedure identical to that discussed in DS08; the local S/N in the data is defined as the mean flux divided by the standard deviation in continuum regions after the data have been smoothed to the resolution element. The S/N(λ) is then modified by setting regions with strong IGM, Galactic, and intrinsic absorption systems and instrumental features equal to zero. W_min,Lyα(z) is calculated from the S/N(λ) vector. The path length Δz(W_min) is then the sum of pixels where W>W_min. As in DS08, we omit regions within 500 km s⁻¹ of the Galaxy and within 1500 km s⁻¹ of the AGN to eliminate (most) absorbers intrinsic to either the AGN or the Local Group. Cosmologically corrected path length ΔX is calculated in an entirely analogous manner, using dX ≡ (1 + z)² [Ω_m(1 + z)³ + Ω_Λ]^−1/2dz. Throughout this paper, we assume a flat (Ω_m, Λ) cosmology with $H_0=(70\,\rm km\,s^{-1} Mpc^{-1})\,{\it h}_{70}$, Ω_m = 0.261, Ω_Λ = 0.716, and Ω_b = 0.0455 h⁻²₇₀ (Spergel et al. 2007). The maximum path lengths available in each sight line for absorbers of any strength are listed in Table 1. The total path length surveyed in all seven sight lines is Δz_tot = 2.193 (ΔX_tot = 1.773).

For the numerator of $(d{\cal N}/dz)_{\rm BLA}$, we have several options to choose from, depending how much faith we place in our BLA designations. The most skeptical view accepts only our probable sample, which with one-sided Poisson statistics gives ${\cal N}_{\rm BLA}=15^{+5}_{-4}$ and $d{\cal N}/dz=7\pm 2$. A more inclusive view includes both the probable and possible groups: ${\cal N}_{\rm BLA}=63^{+9}_{-8}$ and $d{\cal N}/dz=29\pm 4$. Since ∼ 50% of the BLAs survive the statistical correction process described in Section 3.2, we adopt instead an intermediate census: all of the probable BLAs and 50% of the possible sample: ${\cal N}_{\rm BLA}=39^{+7}_{-6}$. Given the uncertainties surrounding BLA identification, we believe pure Poisson uncertainties are far too optimistic. We adopt the skeptical and inclusive censuses as our lower and upper bounds: ${\cal N}_{\rm BLA}=39\pm 24$, $(d{\cal N}/dz)_{\rm BLA}=18\pm 11$ or, in comoving coordinates, $(d{\cal N}/dX)_{\rm BLA}=22\pm 14$. Despite the corrections above, uncertainties in path length are small (≲ 10%), so we ignore errors in the denominator.

Because both BLAs and highly ionized metal lines (O vi, N v, Ne viii, etc.) are thought to trace WHIM gas, it is instructive to look at the overlap in these two samples. O vi has by far the best detection statistics of any FUV intergalactic metal line, with ∼100 detections in the low-z IGM surveys (Danforth & Shull 2008; Tripp et al. 2008; Thom & Chen 2008). In Stocke et al. (2007), we used $\log \,N_{\rm O\,\mathsc {vi}}\ge 13.2$ and $\log \,N_{\rm O\,\mathsc {vi}}<13.2$ as the threshold between a good O vi detection and a reliable nondetection based on the distributions of observed detections and 4σ upper limits. Using the same threshold, we find that four probable BLAs show O vi absorption, while six show O vi nondetections; a detection rate of ∼ 40%. For the larger sample of probable-plus-possible BLAs, the numbers rise to 8 detections and 33 nondetections for an O vi detection rate of ∼ 20%. Using the same criteria, the large DS08 survey (∼ 650 H i systems) features 69 O vi detections and 293 nondetections (19%). However, this sample mixes broad with narrow H i lines. If we instead define a control sample of non-BLAs as all DS08 H i systems with $b_{\rm H\,\mathsc {i}}<40$ km s⁻¹ (516 systems), there are 47 O vi detections and 245 nondetections (16% detection rate, half that of our probable BLA sample). The O vi detection rate is slightly lower (14%) if we use the DS08 b_LW measurements to define the non-BLA control group.

The coincidence of probable BLAs and O vi detections (40%) is several times higher than in the larger, narrow Lyα absorber sample (∼ 15%). This suggests that BLAs and O vi are tracing the same material, but the small number of systems in the BLA sample reduces the significance. It is worth noting that the 40% O vi detection rate is a bit higher than the ∼ 25% fraction of H i systems that show metal absorption in any ion reported by DS08. This suggests that BLAs are accurately tracing WHIM irrespective of metal enrichment. Unfortunately, the detection statistics for other ions are too poor to draw any conclusions.

One of the strongest potential advantages of detecting the WHIM using BLAs is that these absorbers are not affected by the metallicity of the gas, so that even metal-free WHIM can be detected. We would expect these low-metallicity and metal-free absorbers to be found in regions far from galaxies, unlike the O vi absorbers which are typically found within ∼ 800 kpc of the nearest L* galaxy (Stocke et al. 2006; Wakker & Savage 2009) and even closer to sub-L* galaxies (Stocke et al. 2006). Unfortunately, only eight of the BLAs reported here are found in sky regions surveyed for galaxy redshifts complete to L* or below; these are listed in Table 3 in increasing order of galaxy separation. Dividing this sample into O vi detections and nondetections at a consistent level of log $N_{\rm O\,\mathsc {vi}}=13.2$ (DS08), we find nearest galaxies at 0.75–2.9 h⁻¹₇₀ Mpc for the O vi nondetections and 0.06–0.49 h⁻¹₇₀ Mpc for the detections. Therefore, we find some evidence that BLAs are tracing WHIM gas more remote from galaxies than by using O vi absorption as a WHIM tracer.

Table 3. Galaxy–BLA Relationship in Well-surveyed Fields

AGN	zabs	$b_{\rm H\,\mathsc {i}}$a	BLA	log $N_{\rm O\,\mathsc {vi}}$b	d
		(km s−1)	Classification	(cm−2)	(Mpc)
PG 1259+593	0.00229	44 COG	A	13.7:c	0.06
PKS 0405−123	0.16678	75: LW	B	14.1 ± 0.2	0.11
PKS 0405−123	0.09659	70: LW	B	13.7 ± 0.2	0.27
PKS 0405−123	0.08139	53 LW	A	13.3 ± 0.2	0.49
PG 1116+215	0.06072	54 LW	B	<13.07	0.75
PG 1116+215	0.09279	100: LW	A	<13.07	1.3
PG 1116+215	0.01635	51 LW	B	<13.05	2.0
PG 1116+215	0.08587	53 LW	B	<13.04	2.9

Notes. ^aConsensus b-value from Table 2. ^bO vi column density from DS08. ^cDetection from Richter et al. (2004) based on nighttime-only FUSE data. Formal significance level is low (< 3σ), however, absorption appears over an exceptionally broad velocity range (−110 km s⁻¹ < v < +220 km s⁻¹).

Download table as:  ASCIITypeset image

There is one slightly controversial absorber that we have counted as an O vi detection in the above accounting: the z = 0.00229 Lyα absorber toward PG 1259+593 is 60 kpc from the 14h mag edge-on late-type spiral galaxy UGC 8146. DS08 report this absorber as an O vi nondetection according to their 4σ detection threshold, but Richter et al. (2004) report a low-significance, very broad O vi detection at $N_{\rm O\,\mathsc {vi}}=5\times 10^{13}$ cm⁻² based on night-only FUSE data. We thus include this as an O vi detection.

The mass fraction of the local universe traced by broad Lyα absorbers can be determined by dividing the total hydrogen column density by the total observed path length

$$\begin{equation} \Omega _{\rm BLA}=\biggl (\frac{\mu \,m_H\,H_0}{c\,\rho _{\rm crit}}\biggr)\,\frac{\sum N_{\rm H}}{\sum \Delta X}. \end{equation} \tag{ 6 }$$

Since the vast majority of IGM hydrogen is ionized, total hydrogen column density can be approximated for any given absorber by $N_{\rm H}\approx N_{\rm H\,\mathsc {i}}/f_{\rm H\,\mathsc {i}}(T)$. Neutral hydrogen column $N_{\rm H\,\mathsc {i}}$ can be measured directly in most cases. However, the hydrogen neutral fraction $f_{\rm H\,\mathsc {i}}$ is determined by both photoionization from the metagalactic radiation field and ionization due to (thermal) electron collisions. At z ≈ 0, the ionizing background produces an H i photoionization rate $\Gamma _H=\rm 3.2^{+2.0}_{-1.2}\times 10^{-14}\, s^{-1}$ as derived in Shull et al. (1999) from populations and radiative transfer calculations of Seyferts, QSOs, and starbursts. For electron impact, the hydrogen ionization rate can be approximated as

$$\begin{equation} C_i(T)=(5.83\times 10^{-11}\,{\rm cm^3\,s^{-1}})\,\frac{T^{1/2} \exp [-1.58\times 10^5/T]}{1 + 1.58\times 10^6/T}, \end{equation} \tag{ 7 }$$

where 1.58 × 10⁵ K = (I_H/k_B) = (13.6 eV)/k_B. The critical density where collisional ionization equals photoionization is then (n_e)_crit = Γ_H/C_i(T). For borderline WHIM temperatures (T = 10⁵ K), the ionization rate is C_i = 3.6 × 10⁻⁹ cm³ s⁻¹ and the critical density is (n_e)_crit = 8.9 × 10⁻⁶ cm⁻³. However, collisional ionization becomes more and more dominant at higher temperatures, and at log  T = 5.5(6.0), C_i(T) = 1.7(3.1) × 10⁻⁸ cm³ s⁻¹ and (n_e)_crit = 1.9(1.0) × 10⁻⁶ cm⁻³ corresponding to overdensities of δ < 10 at z ∼ 0.

Photoionization is thus an important consideration at low temperatures (k T ≪ 1 Ryd), but at temperatures near or above the O vi peak in CIE (log  T = 5.5–6.0), we have $(n\rm _e)_{\rm crit}=few\times 10^{-6}\,{\rm cm}^{-3}$. At those low densities, in photoionization equilibrium, $f_{\rm H\,\mathsc {i}}=n_e\,(\alpha _H)/(\Gamma _H)\sim (1.3\times 10^{-5}) [n_e/10^{-6}\,{\rm cm}^{-3}]$ we adopt $\Gamma _H = 3.2\times 10^{-14}\,\rm s^{-1}$ and $\alpha _H=4.1\times 10^{-13}\,\rm cm^3\,s^{-1}$ (case A at 10⁴ K). If n_e ≈ (n_e)_crit, collisional ionization could double the ionization rate and halve $f_{{\rm H}\,\mathsc {i}}$ in the formula above (0.65 × 10⁻⁵). A BLA with n_e = (n_e)_crit and $N_{{\rm H}\,\mathsc {i}}=3\times 10^{13}\,{\rm cm}^{-2}$ would have N_H ∼ 5 × 10¹⁸ cm⁻² and line-of-sight dimension $L_{\rm BLA}=N_{\rm H}/n_{\rm H}\sim 5\times 10^{24}\,\rm cm= 1.5$ Mpc. An unvirialized cloud this size would exhibit a 100 km s⁻¹ broadening of its Lyα line width due to Hubble flow from one side to another.

We adopt $f_{\rm H\,\mathsc {i}}(T)$ values derived from a set of CLOUDY simulations (solid curve in Figure 1) featuring both collisional and photoionization (log  U = −2, typical of the low-z IGM) as the most valid approximation to neutral fraction. At WHIM temperatures, this closely follows the CIE neutral fraction, but diverges quickly at T < 10⁵ K. The model ionization parameter U = n_γ/n_H = 0.01 is typical of the low-z IGM, but large changes in the model ionizing field will produce only small changes in $f_{\rm H\,\mathsc {i}}(T)$ at WHIM temperatures. Figure 1 shows model curves for $f_{\rm H\,\mathsc {i}}(T,U)$ based on photo-thermal CLOUDY models with log  U = −2 and log  U = −1; they differ by ∼ 0.3 dex at T ≈ 10⁵ K, but by only ∼ 0.1 dex at T ≳ 10^5.5 K.

Path length $\Delta X(N_{\rm H\,\mathsc {i}})$ for each sight line is calculated as described in Section 3.1. Of the 63 probable and possible BLA candidates, most are strong enough that $\Delta X(N_{\rm H\,\mathsc {i}})\approx \Delta X_{\rm max}$. However the survey of the weaker BLAs is only ∼ 80% complete. We correct for completeness by dividing column density $N_{\rm H\,\mathsc {i}}$ by the corresponding completeness in their respective data (all correction factors were between 0.8 and 1.0). The BLA mass fraction is calculated by modifying Equation (6) to

$$\begin{equation} \Omega _{\rm BLA}=\biggl (\frac{\mu \,m_H\,H_0}{c\,\rho _{\rm crit}}\biggr) \frac{\sum _{i,j} \bigl (f_{{\rm H}\,\mathsc {i}}^{-1}\,N_{\rm H\,\mathsc {i}}\bigr)_{i,j}\,\bigl (\frac{\Delta X_{{\rm max},j}}{\Delta X_{i,j}}\bigr)}{\sum _j \Delta X_{{\rm max},j}}. \end{equation} \tag{ 8 }$$

Given the uncertain relationship between b_obs and b_T for any given absorber, we apply a statistical procedure to better ascertain thermal line widths. First, we assume that b_COG = b_T where available. For the remainder, we perform a Monte Carlo simulation, correcting each observed probable and possible BLA line width as follows. Each candidate is "corrected" using a randomly selected b_COG/b_LW ratio from the 138 absorbers in DS08 with well-determined b_COG (Figure 2). Each BLA candidate is simulated 10⁴ times and the resulting distribution of b_cor is shown in shown in Figure 8. The survival fraction of an individual BLA candidate (i.e., b_cor ⩾ 40 km s⁻¹) is (62 ± 5)% although this includes seven b_COG ⩾ 40 km s⁻¹ absorbers which are not corrected. Roughly 50% of the non-COG BLA candidates survive the correction process.

Zoom In Zoom Out Reset image size

The corrected line widths are used to calculate Ω_BLA for each Monte Carlo simulation. The gas temperature is assumed to be T = 60 b²_cor K (b in km s⁻¹). Neutral fraction as a function of temperature $f_{\rm H\,\mathsc {i}}(T)$ is determined from a CLOUDY simulation as discussed above, and the total hydrogen column density is calculated $N_{\rm H}=N_{\rm H\,\mathsc {i}}/f_{\rm H\,\mathsc {i}}$ for each absorber. Summing over all BLAs, we derive Ω_BLA according to Equation (8). The full Ω_BLA distribution for 10⁴ simulations (Figure 9) shows a median and ±1σ value of Ω_BLA = 6.3^+1.1_−0.8 × 10⁻³ h⁻¹₇₀ or a baryon fraction of Ω_BLA/Ω_b = 14⁺³₋₂%. Varying some of these assumptions changes Ω_BLA as discussed below.

Zoom In Zoom Out Reset image size

4.1.1. Systematic Uncertainties and Trends in Ω_BLA

Previous BLA studies have approached the issues of line identification, line width, and $f_{\rm H\,\mathsc {i}}$ in different ways and most have used some or all of the same data sets studied here. It is instructive to look at the assumptions in and results from these studies to see how different systematics affect $d{\cal N}/dz$ and Ω_BLA. All of these values are shown in Figure 9 in comparison to our results.

Richter et al. (2004) study ∼ 9 BLAs in the PG 1259+593 sight line. They find $(d{\cal N}/dz)_{\rm BLA}\approx 23$ and calculate $f_{\rm H\,\mathsc {i}}$ in CIE from inferred temperature. They note that b_LW overestimates b_T and thus quote only an upper limit Ω_BLA ⩽ 3.3 × 10⁻³ h⁻¹₇₀, roughly half of our result. However, this is a single sight line and the difference may be explainable by cosmic variance.

Richter et al. (2006) analyze H 1821+643 and PG 0953+415 and bring in the results from PG 1259+593 (Richter et al. 2004) and PG 1116+215 (Sembach et al. 2004). They find 20 and 49 BLAs in their "secure" and total samples, with $(d{\cal N}/dz)_{\rm BLA}=22$ and $(d{\cal N}/dz)_{\rm BLA}=53$, respectively. Assuming CIE, they find Ω_BLA>2.7 × 10⁻³(3.8 × 10⁻³) h⁻¹₇₀, but the value becomes unphysically large (> Ω_b) if CIE + photoionization is assumed (dashed curve in Figure 1).

Lehner et al. (2007) compile data from seven sight lines including four from previous work (PG 1259+593 Richter et al. (2004), HE 0226−4110 (Lehner et al. 2006), HS 0624+6907 (Aracil et al. 2006b), and PG 1116+215 (Sembach et al. 2004)), two from as-yet-unpublished private communication (PG 0953+415, T. M. Tripp et al. 2010, in preparation and H 1821+643, K. R. Sembach et al. 2010, in preparation), a detailed reanalysis of PKS 0405−123. They perform quality and column-density cuts on 341 Lyα lines and end up with ∼ 60 BLAs for a density $(d{\cal N}/dz)_{\rm BLA}=30\, \pm\, 4$. To deal with blended components and non-thermal broadening, L07 randomly eliminate 1/3 of their BLAs with 40 < b < 65 km s⁻¹ (all absorbers with b>65 km s⁻¹ are kept). Non-thermal broadening is assumed to be ∼ 10% of the total line width (Richter et al. 2006), so temperatures are calculated assuming b_T ≈ 0.9 b_obs. They calculate Ω_BLA assuming both CIE and Richter's CIE + photoionization parameterization and find Ω_BLA = 3.6 × 10⁻³ h⁻¹₇₀ and 9.1 × 10⁻³ h⁻¹₇₀, respectively, or Ω_BLA/Ω_b = 8% and 20%, for each case.

At higher redshift, Prause et al. (2007) observed BLAs in the spectra of five 1.34 < z < 1.94 AGNs using both the near-UV STIS/E230M grating on HST and the ground-based UV Echelle Spectrograph at the ESO Very Large Telescope. In total, they found 9 good BLA candidates and an additional 29 tentative cases in the redshift range 0.9 ≲ z ≲ 1.9. They derive a value of Ω_BLA = (2.2 ± 0.1) × 10⁻³ h⁻¹₇₀ for the nine good BLA candidates and Ω_BLA = (14 ± 2) × 10⁻³ h⁻¹₇₀ for the entire sample.

4.2.1. Broad Absorbers in Cosmic Origins Spectrograph Observations

Late in the analysis process, we obtained observations of PKS 0405−123 by the HST/COS (J. Green et al. 2010, in preparation; S. Osterman et al. 2010, in preparation) as part of its public Early Release Observations (ERO) program. These data were obtained in the G130M grating ($\rm 1132\ \AA<\lambda <1468$ Å) with a nominal resolution R∼ 20,000 (Δv = 15 km s⁻¹). Seven orbits in total (∼ 17 ks) were devoted to ERO observations, with three taken prior to when an accurate focal alignment was achieved and four afterward. A close examination of the data shows no difference in line profiles for any of the interstellar medium (ISM) or IGM lines of interest, so observations from all seven orbits were aligned and co-added.

The resulting spectrum is of exquisite quality with S/N ⩾40 per nominal seven-pixel resolution element at most locations. Owing to the differences in detector technology and the different grating positions used in the observations, the COS data are free of much of the fixed-pattern noise that plagues the corresponding STIS/E140M observations. This, coupled with the very high S/N, makes COS ideal for verifying the BLA candidates discussed in this paper.

Eighteen of the BLA candidates toward PKS 0405−123 measured in STIS/E140M data are also covered in the COS observations. Five of these cases are blended components of strong absorption systems, which are difficult to confirm or refute, but the COS data are consistent with the STIS observations. Of the 13 weaker absorption lines, seven show a good match between STIS and COS data. However, the STIS-measured line profile is substantially different in four cases and missing altogether in another three. Figure 10 shows several examples where BLA candidates can be either confirmed or refuted based on COS observations.

Zoom In Zoom Out Reset image size

This is a good demonstration of the importance of fully understanding the instrumental effects of a particular spectrograph. While the COS data are not free of fixed-pattern noise, it is independent of that in the STIS observations. Furthermore, it is evidence that the ∼ 10× sensitivity increase of COS over STIS will revolutionize the study broad Lyα absorbers, both through higher S/N and sensitivity to numerous, fainter targets. Since this work relies on a uniform analysis of consistent data sets by independent groups, we do not change any of our BLA designations in light of new, higher-quality data from COS. However, we note the results of our COS absorber verification where possible in the individual absorber comments in the Appendix.

z = 0.06083 (B). Strong line detected in Lyα,β (W_Lyα = 557 ± 20 mÅ, W_Lyβ = 313 ± 113 mÅ) with suggestion of broad wing on blue side and possible component structure. No metal lines to clarify component structure. COG for full system gives b_COG ∼ 36 km s⁻¹ with large uncertainties. We list this as a possible BLA since there the broad wing on the blue side suggests a broad component.

z = 0.16339 (B). Very strong Lyα system. Lyβ is blended with Galactic Si ii absorption, but appears narrow. Lyγ absorption is present but weak. COG solution does not converge and there are signs of multiple components. No metal lines to illustrate component structure.

z = 0.20700 (B). Saturated, broad Lyα profile with possible weak, broad component seen in the wings. Lyβ,γ show asymmetric profiles. Strong O vi, C iii, Si iii (DS08), and Ne viii (Savage et al. 2005) detections as well as possible S iv. We adopt the Savage et al. (2005) measurements for the BLA component, but list the system only as a possible BLA since the component structure is ambiguous.

z = 0.30930 (B). Strong, broad system with probable component structure in Lyα,β,γ: W_Lyα = 388 ± 7 mÅ, W_Lyβ = 130 ± 3 mÅ, W_Lyγ = 33 ± 3 mÅ. Multi-valued COG solution: b_COG = 34 ± 2 km s⁻¹ (α, β) or b_COG = 44 ± 6 km s⁻¹ (α, γ). Due to this ambiguity and the likely component structure, we list this as a possible BLA. No metal lines.

z = 0.38420 (C). Strong system with ambiguous component structure in Lyα. However, Lyβ profile shows two clear components. Two or three components seen in Lyβ profile despite poor S/N. COG on stronger, red, component gives b_COG ∼ 31 km s⁻¹.

z = 0.39641 (B), z = 0.39890 (B), z = 0.40034 (B). Possible continuum ripples shown in Figure 5 and discussed in the text. In particular, a very similar feature appears in the data at 1702.5 Å in the PG 0953+414 data at z_abs>z_AGN which looks similar to the z = 0.40034 feature toward HE 0223−4110, however, other spectra at higher S/N do not show this feature.

z = 0.03196 (B). Broad Lyα profile (W_Lyα = 108 ± 22 mÅ, b_LW = 57 ± 8 km s⁻¹), but weak, narrow Lyβ (W_Lyβ = 50 ± 6 mÅ, b_LW = 35 ± 4 km s⁻¹). COG is poorly constrained: b_COG>10 km s⁻¹. Possible BLA based on Lyα alone. COS observations are consistent with the STIS/E140M data.

z = 0.05896 (C). Noisy data suggest two weak, narrow components each with b ∼ 20 km s⁻¹. COS spectrum confirms the two components with b_LW = 23 and 27 km s⁻¹ are not BLAs.

z = 0.07218 (C). Very marginal feature which is not clearly Lyα. Identified as C iii z = 0.3340 by DS08. Absorption feature confirmed in COS data as real with b_LW = 30 ± 3 km s⁻¹.

z = 0.07523 (B). Very weak detection measured by both W06 (b_LW = 56 km s⁻¹) and L07 (b_LW = 48 ± 20 km s⁻¹). High-S/N COS observations show no absorption at this position (W < 11 mÅ).

z = 0.08139 (A). Moderately strong system with b_LW ∼ 53 km s⁻¹, W_Lyα = 261 ± 16 mÅ. Lyβ shows W = 90 ± 8 mÅ but is blended with Ly z = 0.18269 (W_Lyα ∼ 670 mÅ). If this entire absorption feature is taken as Lyβ z = 0.08139, then b_COG>22 km s⁻¹. However, if only part of absorption is taken as Lyβ (W_Lyβ ∼ 45 ± 10 mÅ), the COG line width is large but uncertain: b_COG = 74 ±  30 km s⁻¹. Given this ambiguity and the broad, clean Lyα profile, we list the absorber as a probable BLA.

z = 0.09659 (B). Strong Lyα system (W_Lyα ∼ 500 mÅ) with excess in blue wing indicating possible broad component. DS08 find b_COG = 36 ± 4 km s⁻¹ and W06 find b_LW = 37 ± 1 km s⁻¹ for the entire system. However, L07 fit a broad component to the system and we take this as a possible BLA due to the ambiguous nature of the fit. DS08 report O vi detected in both lines of the doublet associated with the system, though it is unclear to which component it may correspond. COS observations show consistent line profile, but BLA component cannot be verified.

z = 0.10298 (C). Very marginal feature which is not obviously a single absorber. W06 measure b_LW = 60 ± 14 km s⁻¹. COS data show a clear blend of two absorbers b_LW = 27 ± 5 and 42 ± 3 km s⁻¹, thus no BLA is confirmed in this system.

z = 0.10419 (C). Very marginal detection not confirmed by DS08. COS observations show no absorption at this position (W < 10 mÅ).

z = 0.13102 (B). Moderate absorption feature with broad wing on red edge suggesting possible BLA component. COS observations show absorption at this wavelength (b_LW = 35 ± 1 km s⁻¹, log  N = 13.18 ± 0.01), significantly narrower than that observed by STIS.

z = 0.13377 (B). Noisy data with moderately broad absorption. COS observations show a clear absorber at this location (b_LW = 53 ± 2 km s⁻¹, log  N = 13.10 ± 0.02) but significantly shallower and broader than the STIS profile.

z = 0.13646 (C). Marginal line with likely components. Single-component fit from W06 gives b_LW = 49 ± 8 km s⁻¹. COS confirms double component structure, b_LW = 23 ± 4, 26 ± 2 km s⁻¹.

z = 0.13924 (C). No detection in L07. Marginal feature in DS08 and W06. Not clearly a single component even if real. COS observations show two absorbers offset in velocity, but nothing corresponding to the reported line profile.

z = 0.15304 (C). Strong, broad line (W_Lyα = 263 ± 10 mÅ) identified as Lyα by L07, but corresponding Lyβ nondetection (W_Lyβ < 24 mÅ) is inconsistent with a single Lyα absorber. Unknown line identification. COS confirms line profile.

z = 0.16121 (C). Strong absorber with several clear components. Strong metal lines likely make up the flanking components leaving little room for a broad Lyα core. COS confirms general profile. Thus, COS does not confirm this BLA.

z = 0.16678 (B) and z = 0.16714 (C). Strong Lyα complex (W_Lyα ∼ 650 mÅ) with an asymmetry on the blue wing. Line can be fitted with a broad and narrow component, but decomposition is suspect. Lyβ (W_Lyβ ∼ 450 mÅ) shows no sign of the broad, blue component. Strong system (z = 0.16714) shows obvious component structure in various metal lines. The system shows strong O vi and N v absorption (DS08), but it is unclear with which component it is associated. COS observations show a consistent line profile, but cannot confirm or deny the presence of a BLA component.

z = 0.17876 (B). Asymmetric Lyα line with very marginal Lyβ detection. COG analysis by DS08 gives b_COG = 18⁺⁴⁴₋₇ km s⁻¹. Apparent optical depth (AOD) line fits to Lyα line give much broader profile (b_LW ∼ 60 km s⁻¹; W06 measures b_LW = 58 ± 6 km s⁻¹). Well fit by an offset broad+narrow pair (b_LW = 57, 14 km s⁻¹), but inconclusive. We take the AOD measurements as concensus in this case. COS observations confirm the general profile shape of this absorber and consistent line measurements.

z = 0.18269 (B). Strong H i system (W_Lyα ∼ 690 mÅ) with at least two components, possibly more. Lyβ is blended with Galactic Lyα. Lyγ shows strong, moderately asymmetric profile (W_Lyγ = 139 ± 11 mÅ). COG gives b_COG = 49⁺¹¹₋₇ km s⁻¹, but probable component structure relegates this to a possible BLA. System shows several O vi components. COS observations confirm the line profile of this strong absorber, but cannot confirm or deny the presence of a BLA.

z = 0.19086 (B). Weak, noisy absorber with possible component structure. COS observations do not show an absorption line at this position (W_r < 22 mÅ).

The COS ERO spectral coverage (1132 < λ < 1468) ends at this maximum redshift.

z = 0.24513 (B) and z = 0.24553 (X). Pair of well-separated Lyα absorbers. System at z = 0.24513 appears narrow (b_LW = 30 ± 5 km s⁻¹) but both L07 and W06 measure b ≈ 55 km s⁻¹. We adopt a consensus b and N intermediate between the two measurements to account for possible continuum and reduction differences.

z = 0.32500 (C). Marginal detection, however, there is a correlation between features in Lyα,β. Probably multiple components. DS08 report b_COG = 21 km s⁻¹, however, we find this to be more reasonable as a lower limit and the COG solution does not fit the absorption profile very well. W06 report b_LW = 81 ± 11 km s⁻¹. Higher S/N data on this absorber should prove interesting. Given the marginal nature of this feature and the likely component structure, we list it as a non-BLA.

z = 0.35092 (B), z = 0.35149 (X). Strong Lyα system flanked by Si iii z = 0.36079 and Lyα z = 0.35149. Lyα, β COG gives b_COG = 56 ± 15 km s⁻¹ (DS08), but Lyα,γ COG gives b_COG = 27 ± 5 and seems to better match the profile. W06 and L07 measure b_Lyα = 38 ± 2 km s⁻¹ and b_Lyα = 40 ± 8 km s⁻¹, respectively. Both Lyα and Lyβ show asymmetric profiles suggesting multiple components. We accept the Lyα measurements as consensus in this case and list the system as a possible BLA.

z = 0.36150 (C) and z = 0.36079 (X). Strong, blended Lyα systems with an ambiguous component structure and noisy data. Two-component fit gives b_LW = 75 ± 10, 39 ± 5 km s⁻¹ for the z = 0.36150 and z = 0.36079 components, respectively. The z = 0.36150 component, however, does not appear in Lyβ (W_Lyβ ⩽ 9 mÅ); based on the Lyα component, the expected Lyβ line should be W_Lyβ ∼ 50 mÅ. Meanwhile, a Lyβ, γ COG fit to the z = 0.36079 component (ignoring the noisy, possibly blended Lyα profile) gives b_COG = 25 ± 3 km s⁻¹, log  N = 15.18 ± 0.13 cm⁻². We conclude that neither system is a BLA.

z = 0.40886 (B). Strong Lyα system (W_Lyα = 430 ± 10 mÅ) at the long-wavelength, noisy end of the STIS/E140M range. Lyβ,γ also present (W_Lyβ = 126 ±  15 mÅ, W_Lyγ = 68 ±  10 mÅ), both with asymmetric profiles. Lyα,β and Lyα,γ COG solutions inconsistent but generally b_COG ∼ 40 km s⁻¹. Possible BLA, but probable component structure.

z = 0.04116 (C). Weak feature consistent with Si iii z = 0.06346, a system in which many other ionic lines are also detected.

z = 0.05437 (B), z = 0.05515 (B), and z = 0.05484 (C). Strong Lyα system (W_Lyα ∼ 450 mÅ) with several plausible BLA subcomponents. The z = 0.05437 line appears as a very marginal, noisy absorption on the blue wing while the z = 0.05515 component is an excess on the red side of the main absorption line. The strong central component at z = 0.05484 is fitted well by a b_LW ∼ 35 km s⁻¹ component despite the b_COG = 45⁺²⁵₋₉ km s⁻¹ value reported in DS08. The narrow feature at 1283.15 Å can be unambiguously identified as Si iii z = 0.0635. The BLA nature of these lines relies on the details of the component structure which is quite ambiguous and difficult to disentangle.

z = 0.06346 (C). Strong Lyα system with measured b_LW>40 km s⁻¹ but b_COG < 40 km s⁻¹ with O vi detected in both lines of the doublet (DS08). H i profile looks symmetric, but there are two clear components in Si iv, C iv, Si iii, and S iii.

z = 0.13597 (B). Weak double profile. Feature at 1381.4 Å is Lyβ z = 0.3468 but broader feature at 1381.0 Å is a plausible BLA with low-significance O vi detections at both 1032 and 1038 Å. Possible BLA based on the measured line though the large error bars and O vi detection could argue for this being a probable BLA.

z = 0.30994 (B). Broad, weak absorption feature identified as Lyα by both L07 and DS08. However, considerable continuum uncertainty yields highly incompatible b, N solutions. Possible, low-significance O vi detected as well. In any case, profile is asymmetrical and not obviously a single component.

z = 0.31045 (C), z = 0.31088 (C). Broad, asymmetric features which are not obviously real. Possible continuum differences.

z = 0.31790 (B). Moderate, narrow Lyα line with asymmetry on the red wing and good O vi detections in both lines of the doublet. Possible multiphase system and/or continuum fit uncertainties. Possible BLA.

z = 0.32089 (C). Strong, moderately narrow line is a poor fit to DS08 COG solution (b_COG = 44⁺¹³₋₁₄ km s⁻¹). Lyα absorption is anomalously large given observed Lyβ,γ, however, there are no other obvious contributors to the Lyα line strength. We confirm the b_LW = 31 ± 1 km s⁻¹ solution of L07 for this absorber.

z = 0.33976 (A). Strong, broad system with no obvious components and detections in Lyα,β,γ. Good COG solution gives b_COG = 41 ± 6 km s⁻¹. O vi detected in both lines of the doublet (DS08).

z = 0.04382 (C). Reported as O vi λ1032 at z = 0.22974 by Tripp et al. (2008). Interestingly, this is a "blind" O vi system with little or no accompanying Lyα absorption (W_Lyα < 7 mÅ).

z = 0.05879 (B). Narrow line with a possible excess in line wings suggesting a multiphase system. Very ambiguous decomposition. We take the L07 values as the consensus.

z = 0.12784 (C). Claimed broad blue wing on a narrower Lyα system. We see no good evidence for this system in our reduction of the data.

z = 0.17985 (B). Weak single or double absorption. Potentially fit as single BLA or two narrow components (b_LW ∼ 20).

z = 0.19126 (B). Asymmetric Lyα profile with uncertain decomposition. Formal AOD line width is b_LW ≈ 66 km s⁻¹ , however, two components (b_LW = 40, 32 km s⁻¹) provides a better fit. We adopt b_LW = 40 km s⁻¹ and log  N = 13.3 cm⁻² as consensus values.

z = 0.19361 (C). Strong narrow Lyα system with very weak Lyβ,γ counterparts. DS08 report a Lyα,β solution (b_COG = 24⁺⁴₋₃ km s⁻¹, log  N = 14.15 ± 0.04 cm⁻²) which is stronger than the data can easily support. We measure b_LW = 37 ± 2 km s⁻¹ instead and adopt that value here.

z = 0.08587 (B), z = 0.08632. Pair of very shallow, ripples in the continuum with b_LW ∼ 55 and b_LW ∼ 34 km s⁻¹, respectively, both with considerable uncertainty. Broader feature is possible BLA.

z = 0.09279 (A). Very weak, broad system blended with two weak, narrow features (possible Galactic C i). While profile looks questionable, it is more obvious when looking at a broader wavelength range. Eliminating the two sharp features gives a very plausible BLA.

z = 0.13370 (A). BLA with a low-significance O vi λ1032 detection ($\log \,N_{\rm O\,\mathsc {vi}}\approx 13.2\pm 0.3$ cm⁻²). Potentially fit as two components (λ ≈ 1378.3, 1378.7) but b_LW ∼ 50 km s⁻¹ even in this case.

z = 0.00229 (A). Lyα line appears on the edge of the broad Galactic Lyα trough and Lyβ is blended with an O i airglow line in FUSE data. Richter et al. (2004) used night-only data to get a good Lyβ measurement and constrains b_COG = 44⁺⁹₋₄ km s⁻¹ which we adopt here.

z = 0.04606 (C). Very strong system with obvious components in O vi, C iv, Si iii, and Si iv. Richter et al. (2004) report two H i components, both with b_COG < 40.

z = 0.14852 (A). Strong nearly triangular line profile with detections in Lyα,β. COG consistent with BLA b_COG ≈ 57 km s⁻¹ though with large errors. b_Lyβ ∼ 36 km s⁻¹.

z = 0.19573 (B), z = 0.19620 (X). Double absorption profile. Shallow, blue feature at 1453.6 Å is plausible BLA candidate (b_LW = 46 ± 19 km s⁻¹) and was reported as a component to the z = 0.1962 system by Richter et al. (2004). Marginal N v detection.

z = 0.21136 (C) and z = 0.43569. Richter et al. (2004) identify the absorption at 1472.6 Å as Lyβ z = 0.43569 and high-resolution FOS spectra confirm Lyα line at this redshift. Absorption at 1242.5 Å where Lyβ z = 0.21136 would be expected is blended with Lyα z = 0.02217 and unrecoverable. Cannot confirm.

z = 0.11133 (A), z = 0.11167 (X). Double absorber profile with one broad, one narrow system (b_LW = 52, 31 km s⁻¹). L07 measure this as a single system (b_LW = 88 ± 14 km s⁻¹). One BLA likely present.

z = 0.12147 (A), z = 0.12117 (X). Strong and highly asymmetric absorber. Red wing Lyβ is hinted at in the data, but at low significance. System shows b_COG>40 km s⁻¹, but there are clearly multiple components. Weak O vi absorption is reported in the system (Tripp et al. 2001, 2008; Oegerle et al. 2000) aligned with broad Lyα wing, however, DS08 do not confirm this detection.

z = 0.14754 (B) and z = 0.14776 (X). Blended system of two, easily separable components (b_LW ≈ 42, 20 km s⁻¹). Marginal Lyβ detection for both components but COG solution is very poorly constrained (b_COG = 22^+∞₋₁₂ km s⁻¹) for the z = 0.14754 system.

z = 0.21326 (A). Strong system with Lyα,β detections (b_COG = 42⁺⁵₋₄ km s⁻¹) though with some hint of component structure. Strong O vi detection in both lines of the doublet. The entire absorber may not represent hot gas, but there is probably a BLA component.

z = 0.22480 (C). Very strong system (W_Lyα ∼ 800 mÅ) with clear components in higher Lyman lines. O vi detected. Not confirmed by L07, but clearly not a BLA.

z = 0.22616 (B). Moderate absorption feature (W_Lyα = 150 ± 8 mÅ) identified as Lyα by L07. However, corresponding Lyβ nondetection shows W_Lyβ ⩽ 10 mÅ, inconsistent with a single Lyα system. Possible O vi absorption ($\log \,N_{\rm O\,\mathsc {vi}}=13.4\, \pm 0.1$ cm⁻²) in both lines of the doublet offset ∼+60 km s⁻¹ from the purported H i system. Probable multiple components or misidentification.

z = 0.25814 (B). Strongly asymmetric profile suggests a multiphase H i system. Profile is reasonably fitted by two components: b_LW ≈ 56, 17 km s⁻¹. Total line shows W_Lyα = 141 ± 3 mÅ and corresponding weak Lyβ absorption (W_Lyβ ≈ 24 mÅ) gives a COG solution b_COG = 30 ± 7 km s⁻¹ for the combined system. In light of this, the component fits seem reasonable and we list this as a possible BLA.

z = 0.26658 (A). Strong system with O vi detected in both lines of the doublet. Poorly constrained COG gives b_COG = 68 ± 30 km s⁻¹. Individual lines show b_LW = 46 km s⁻¹ and b_LW ∼ 40 km s⁻¹. Probable BLA despite poorly constrained COG.