The C IV forest as a probe of baryon acoustic oscillations
aa r X i v : . [ a s t r o - ph . C O ] O c t Mon. Not. R. Astron. Soc. , 1–6 (x) Printed 25 June 2018 (MN L A TEX style file v2.2)
The C IV Forest as a probe of baryon acoustic oscillations
Matthew M. Pieri ⋆ Institute of Cosmology & Gravitation, University of Portsmouth, Dennis Sciama Building, Portsmouth PO1 3FX, UKAix Marseille Universit´e, CNRS, LAM (Laboratoire d’Astrophysique de Marseille) UMR 7326, 13388, Marseille, France
Accepted xxxx
ABSTRACT
In light of recent successes in measuring baryon acoustic oscillations (BAO) in quasarabsorption using the Lyman a (Ly a ) transition, I explore the possibility of using the 1548 ˚Atransition of triply ionized carbon (C IV ) as a tracer. While the Ly a forest is a more sensitivetracer of intergalactic gas, it is limited by the fact that it can only be measured in the opticalwindow at redshifts z >
2. Quasars are challenging to identify and observe at these high red-shifts, but the C IV forest can be probed down to redshifts z ≈ .
3, taking full advantage of thepeak in the redshift distribution of quasars that can be targeted with high efficiency.I explore the strength of the C IV absorption signal and show that the absorbing populationon the red side of the Ly a emission line is dominated by C IV (and so will dominate over thepotential BAO signal of other metals). As a consequence, I argue that forthcoming surveysmay have a sufficient increase in quasar number density to offset the lower sensitivity of theC IV forest and provide competitive precision using both the C IV autocorrelation and theC IV -quasar cross correlation at h z i ≈ . Key words: large-scale structure of Universe, distance scale, dark energy, intergalacticmedium, quasars: absorption lines, cosmology: observations
Baryon acoustic oscillations (BAO) are imprinted on large-scalestructures and provide a probe of cosmic expansion throughtheir use as a standard ruler (Weinberg et al. 2013 and referencestherein). Since the intergalactic medium is thought to represent theoverwhelming majority of baryons at z > a forest absorbers along the lineof sight to distant quasars (Busca et al. 2013; Slosar et al. 2013;Delubac et al. 2014). They measure the BAO scale with a uncer-tainty of 2% at h z i = . a forest absorption and quasars at thesame redshift (Font-Ribera et al. 2014a), which is more effective asa probe of the angular diameter distance. Together they have provento be a powerful cosmological probe (Delubac et al. 2014).The BAO peak has been observed at z . . < z < .
5. Ly a forest based studies are not possible at z < ⋆ E-mail:[email protected] as Ly a leaves the optical band, but metal forests with longer transi-tion wavelengths allow the possibility of tracing structure at lower-redshifts. Triply ionized carbon (C IV ) provides a useful doubletC IV a ( l IV b ( l IV forest to measure the BAO scale and so provide a new probe ofthe expansion of the universe. This work is set out as follows: inSection 2, the data set used will be described, in Section 3, thestrength of the C IV signal is explored, Section 4 describes testsof the metal population, Section 5 presents a potential survey, andSection 6 explores some observational challenges. I use the Sloan Digital Sky Survey III Data Release 9 (Ahn et al.2012) quasar sample compiled by visual inspection of candidatesas outlined in Pˆaris et al. (2012). Ly a absorption with a redshiftrange 2 < z < . > a and C IV forests as described in Section 3.A broader sample of quasars, including 9523 at low redshifts takenfrom Data Release 9, was used to blindly explore absorption on c (cid:13) x RAS Pieri M. M. et al. the red side of the Ly a emission line corresponding to C IV withredshifts 1 . < z < . > IV and Si IV absorp-tion; the ‘C IV +Si IV band’) and 1400–1520 ˚A (without Si IV ab-sorption; the ‘C IV band’). The boundaries of these bands are con-servatively defined to avoid uncertainly in fitting Ly a , Si IV andC IV emission lines. I use two methods for characterizing the comparable signal strengthbetween Ly a and the C IV forests. The most direct statistical com-parison of the bulk metal signal and the bulk Ly a forest is pre-sented by measurements of the 1D power in both the Ly a for-est and on the red side of the Ly a emission line. Fig. 19 ofPalanque-Delabrouille et al. (2013) shows the power in the tworestframe bands of interest, while fig. 20 shows the 1D power spec-trum measured in the forest at various redshifts. In comparing thesetwo figures one can see that the metal power in these red side bandsis a factor of 5–20 lower than that of the forest power on scalesgreater than that of the doublet ( k < . ( km / s ) − ) at 4000 ˚A, themean wavelength of current BAO measurements from Ly a . Thiscorresponds to z = . a forest or z = . IV for-est. It is also notable when comparing the power spectrum of eachred side band that while there is more power in the C IV +Si IV bandthan C IV band, the difference indicates that the additional Si IV power is no greater than 50% of that present in the C IV band.We can also take a more focused approach by comparing theprimary measured quantity in Ly a forest BAO studies, the flux con-trast, d ( l ) = f ( l ) / ( C ( l ) ¯ F ( z )) −
1, for both the C IV and the Ly a forests, where f ( l ) and C ( l ) are the quasar flux and the absorbedcontinuum level of the quasar respectively in the relevant wave-lengths ranges, F = f / C is the transmitted flux fraction (referredto as ‘flux’ hereafter), and ¯ F ( z ) is the mean flux in the appropriateforest. This requires the C IV to be present at the same redshift asthe Ly a absorption and so is a test of C IV at z > a forest and determin-ing the flux contrast for C IV at the same redshift. The flux contrastof this pixel pair is then calculated by renormalizing locally. Anal-ogous to the locally calibrated pixel search of Pieri et al. (2010a)and the composite spectra approach (Pieri et al. 2010b; P14), thisbrute force method cancels the contribution of contaminating ab-sorption in the local renormalization. The average d CIV is calcu-lated in bins of d Ly a . Fig. 1 compares these quantities for the fullrange of d Ly a . This indicates that the C IV forest signal is between 3and 20% of the Ly a forest signal over most of the dynamic range in d Ly a . Where the Ly a signal is strongest, the proportionate strengthof the C IV signal is also strongest, corresponding to higher densityand potentially higher bias systems. This provides a conservativeassessment of the C IV forest signal strength as it neglects scatter inabsorption strength and implicitly assumes that any inhomogene-ity in C IV absorption strength is stochastic. We shall return to thispoint in Section 5.I conservatively conclude that the signal power in the C IV aforest is a factor of 5–30 lower than that of the Ly a forest at2 < z < . z = .
6, implying that the
Figure 1.
The ratio of the flux contrast magnitudes for the C IV and Ly a forests. The C IV forest is typically 3–20% weaker than the Ly a forest. Thestationary points in this ratio occur as absorbers with mean transmission inthe Ly a forest have excess transmission in the C IV forest. The relative C IV signal is strongest for strong Ly a absorbers. C IV a forest dominates these bands. In the following section I testfor other metal absorbers present in the red side bands in order totest this inference, and explore their potential for adding systematicerrors. A key requirement for a viable probe of BAO is the degree ofcontaminating absorption arising due to other metal absorbers.Unwanted absorption may add both systematic and stochasticnoise to a 3D large scale structure measure (Slosar et al. 2011;Delubac et al. 2014; Bautista et al. in preparation).In order to explore the populations of metal absorbers inthe two red side spectroscopic bands, I stack pixels with flux ( . n − . ) F < ( . n + . ) (where n = , ,...,
9) amongstthe high S/N subset of DR9 including low redshift quasars. I do thisblindly with no attempt to limit the source absorption transition. Ineach case the absorber is assumed to have a rest-frame wavelengthof 1548 ˚A such that the observed wavelength is l o = ( + z ) (so the wavelength is scaled by a factor of 1 / ( + z ) before stack-ing). In the resultant composite spectra, the size of the absorbingpopulations from species such as C IV , Si IV and Mg II can be as-certained using their doublet lines. The line strength ratio arisingfrom the exclusive selection of any of these doublet lines is deter-ministic, and so deviations in the strength of these features in thecomposite spectrum provide a direct measurement of dilution fromother absorbers or noise. For example if C IV a is selected half thetime, then one would expect the flux decrement ( D ≡ − F ) asso-ciated with C IV b to be half that for perfect C IV a selection (with asmall correction for the mean opacity of the metal forest)Fig. 2 shows an example composite spectra of pixels with aflux between 0 . F < .
35. Model profiles are plotted projectingline strengths assuming 100% selection of the line labelled. Hence,for a selected transition l selec and its correlated transition l correl the resultant feature would be present at l correl l CIVa / l selec . Thetop panels of Fig. 2 shows the flux contrast in the C IV band (left)and the C IV +Si IV band (right). In the regions bounded by the grey c (cid:13) x RAS, MNRAS000
35. Model profiles are plotted projectingline strengths assuming 100% selection of the line labelled. Hence,for a selected transition l selec and its correlated transition l correl the resultant feature would be present at l correl l CIVa / l selec . Thetop panels of Fig. 2 shows the flux contrast in the C IV band (left)and the C IV +Si IV band (right). In the regions bounded by the grey c (cid:13) x RAS, MNRAS000 , 1–6 he C IV forest as a probe of BAO Figure 2.
Example composite spectra produced by stacking metal absorption with 0 . F < .
35 blindly on the red side of quasar Ly a emission lines. Theleft-hand panels show stacking in the C IV band and the right-hand panels show stacking of absorption in the C IV +Si IV band. The top panels show the fit(in blue) to the stacked feature (shifted the rest frame wavelength of C IV a and marked with a dashed line). This fit was performed over a region bounded bythe grey vertical lines. In the it middle panels, the absorption due to the selected feature is corrected for using this fit. Projected absorption based on this fitand arising from 100% selection of absorbing transitions are shown as marked. In the it bottom panels, signal due to the selection of C IV doublet absorbers iscorrected for by rescaling the above projection to match the apparent population sizes, and projected absorption due to the selection of Mg II doublet absorptionis shown. In each case the difference between projected and observed absorption provides a measure of the fractional incidence of stacking absorbers from thelabelled transition. vertical lines, a model Gaussian flux decrement profile was fitted tothe feature arising due to absorber selection, shown as a blue line.The middle panels of Fig. 2 show the residual d F when the fittedabsorption profiles have been corrected for in both red side bands(by adding fitted flux decrements to the flux contrast in the top pan-els). In the C IV band, projected doublet line signal associated withperfect selection of both C IV doublet members are shown. In theC IV +Si IV band, this is repeated with the addition of projectionsfor the Si IV doublet again assuming perfect selection. The bottompanels of these figures show the residual d F corrected for the C IV doublet absorption shown in the middle panels and rescaled to theobserved C IV absorber fractions in the composite. The projectedabsorption due to perfect selection of Mg II doublet members isshown here.Using the above procedure, the fraction of doublet lines fromC IV , Mg II , and Si IV were measured. Contributions from other red side metals such as Al II , Al III and Fe II were explored, but no in-dications of their presence in the composite were found. Hence,we shall assume that all selected pixels not identified as one of theabove species arise due to noise.Fig. 3 shows the measured absorber fractions in our sample inboth red side bands derived from this analysis. An uncertainty inthe residual d F of approximately 0.03 is introduced by the imper-fect fitting and removal of absorption and this leads to an uncer-tainly of approximately 0.05 in the absorber fractions that result.As such measurements of Si IV b and Mg II may be viewed as upperlimits. In both red side bands the strongest carbon doublet memberC IV a dominates the metal absorber population, representing 60-70% in the C IV band and 50-70% in the C IV +Si IV depending onthe absorber strength. In all but the strongest absorbers, the nextlargest contribution comes from the weaker member of the C IV c (cid:13) x RAS, MNRAS , 1–6 Pieri M. M. et al. doublet. When combined the C IV doublet make up around 80% ofabsorbers in both bands. a and C IV as structure tracers C IV a is the dominant absorption species in the C IV +Si IV and C IV bands. This constitutes a maximum redshift path of D z = . a BAO measure-ments. The bands described here are somewhat narrow and mightbe expanded given testing, particularly into the Si IV emission lineregion defining a gap between bands.C IV b as the most significant additional absorbing popula-tion may contribute to the BAO measurement signal. The rest-frame wavelengths differ by 0.2%, much smaller than the antici-pated BAO scale precision, hence the C IV b signal can be used toboost the signal (although the effective rest-frame wavelength ofthe combined doublet must be determined mocks to ensure that asystematic error is not introduced). Even a small contribution fromSi IV might not prove to be a significant contaminant as the line-of-sight BAO scale at a given observed wavelength is approximatelyequal in C IV and Si IV for concordance cosmology. Mg II may alsopresent a contaminant, but for concordance cosmology the trans-verse BAO separation differs from that of C IV by a factor of ∼ a absorption strength) is a factor in setting the C IV opacity(Pieri, Schaye & Aguirre 2006; Turner et al. 2014). Furthermore,results from P14 indicate that the strongest Ly a forest absorbers(with d Ly a < .
58) at SDSS resolution and in the absence of noisetypically represent circumgalactic regions. Hence, the C IV forest islikely to be more biased than the Ly a forest, which would increasethe relative clustering amplitude in the C IV autocorrelation.A further potential benefit of using the C IV forest arises dueto its relative dynamic range. The Ly a forest saturates when struc-tures reach an over density as low as 10. As a weaker absorbingpopulation as a whole, the C IV forest rarely saturates and has thepotential to distinguish higher density systems, which may alsolead to a more refined probe of biased systems. I forego a full Fisher matrix projection of cosmological constraintsin light of the uncertainties in the use of the C IV forest as a probe.Instead I use a simple scaling relationship argument. In Section 3,we found that the C IV a forest is a factor of 5–30 weaker than theLy a forest, and since the C IV b forest can be used to boost thesignal this becomes a factor of 4–25. This is consistent with themeasurement that the total 1D power in the C IV forest is a fac-tor of 5–20 lower that the 1D power in the Ly a forest. We maycompare these probes directly for a C IV absorption survey withmean redshift h z i ≈ . a forest surveys h z i = . Figure 3.
The fraction of metal absorbers arising due to the labelled transi-tions in the C IV band (top) and the C IV +Si IV band (bottom). Fractions aremeasured by comparing projected and observed absorption for doublet linesin average composite spectra such as that shown in Figure 2. The fractionof absorbers is presented assuming that all selected pixels without identifiedassociated arise due to stochastic noise (i.e. fractions sum to 1). assuming similar bias, data quality, contaminants, effective vol-ume and path length. Following the formalism of equation 12 ofMcQuinn & White (2011) (see also McDonald & Eisenstein 2007),a decline in signal power must be accompanied by an increase inquasar number density in order to retain equal precision on the mea-surement of the BAO scale.eBOSS (with SDSS-II DR7 and BOSS data) is expected toobtain spectra of 1 . < z < .
15 quasars with approximately fourtimes the number density of BOSS z > .
15 quasars (C. Y`eche,private communication). Quasars with redshifts up to z = . IV absorption below z = a for-est autocorrelation. Therefore, we may conclude that eBOSS willprovide a C IV autocorrelation measurement of BAO with precisionof order that obtained by the DR11 Ly a autocorrelation analysis ofDelubac et al. (2014), assuming all other observing conditions tobe equal. This corresponds to a measurement uncertainty of 2%.The primary science driver for obtaining this high surfacedensity of low redshift quasars is the measurement of BAO usingquasar redshifts. This would probe structure at largely the sameepoch as proposed here, but with different systematic errors and ob- c (cid:13) x RAS, MNRAS000
15 quasars (C. Y`eche,private communication). Quasars with redshifts up to z = . IV absorption below z = a for-est autocorrelation. Therefore, we may conclude that eBOSS willprovide a C IV autocorrelation measurement of BAO with precisionof order that obtained by the DR11 Ly a autocorrelation analysis ofDelubac et al. (2014), assuming all other observing conditions tobe equal. This corresponds to a measurement uncertainty of 2%.The primary science driver for obtaining this high surfacedensity of low redshift quasars is the measurement of BAO usingquasar redshifts. This would probe structure at largely the sameepoch as proposed here, but with different systematic errors and ob- c (cid:13) x RAS, MNRAS000 , 1–6 he C IV forest as a probe of BAO serving biases. Font-Ribera et al. (2014a) demonstrated that muchcan be gained from cross-correlating quasar redshifts with the Ly a IV -quasar autocorrelation has the potential to mea-sure the BAO scale with uncertainty as low as 1%. Renormalizing quasar spectra to correct for the unabsorbed con-tinuum level represents a significant challenge to performing cos-mology in data of this quality. The use of an absorption probe withsignal reduced by an order of magnitude or more might naivelybe presumed to preclude a useful measurement. However, the highincidence rate of Ly a absorption in high redshift studies createslimitations that are not present in the C IV forest. The Ly a forestis sufficiently dense by comparison to the BOSS spectral resolu-tion that it is not possible to identify unabsorbed regions with highconfidence. Hence alternative approaches are typically used such asjoint fits to a stacked mean quasar spectrum and the forest flux prob-ability distribution function (Busca et al. 2013) or the production ofprincipal components based on low redshift quasars that lack thisdense absorption (Lee et al. 2013). The C IV forest is sufficientlysparse that identifying unabsorbed regions in the data will be sub-stantially easier, allowing simple spline fits such as those used here.Mocks will be required to test a variety of systematics in thedata. We may utilize the methods described by Bautista et al (inprep) (see also Font-Ribera, McDonald & Miralda-Escud´e 2012;Delubac et al. 2014), but some additionally refinements are nec-essary in the generation of metal forest mocks. The physics gov-erning the distribution of metals is more complex than the physicsgoverning the distribution of neutral hydrogen. Metals only arise inthe intergalactic medium by virtue of extragalactic mechanical out-flows and it is an open question as to how they are distributed withrespect to faint (typically unobserved) high redshift galaxy popula-tions (e.g. Rahmati & Schaye 2014).Neutral hydrogen is thought to trace large-scale structure tofirst order, but observed metal absorption is far from ubiquitous.As discussed in Section 5, observations indicate that the metal en-richment is non-uniform. Also different metal ionization speciesare thought to trace different subsets of metal enriched gas. Thepatchy nature of the C IV forest may pose a challenge in the pro-duction of mock spectra. However, absorption based BAO mea-surements are thought to be largely insensitive to small scale gasphysics (McQuinn & White 2011) and it is possible to explore arange of metal population scenarios in mocks to test BAO mea-surement and bias is affected.An effect that must be explored are fluctuations in the extra-galactic ionizing UV background. The different ionization poten-tial of C IV compared to H I and the extra ionization states in car-bon complicate the ionization characteristics of C IV . Further po-tential effects may be present if the C IV absorption tends to arisein regions close to galaxies with excess UV levels compared to thebackground. Also inhomogeneity in the C IV fraction may arise onlarge scales due to post helium reionization proximity to quasars. The severity of this effect is somewhat limited (Pieri et al., in prepa-ration), but it is notable that such an effect may impact upon thefeasibility of quasar-absorption cross-correlation using C IV . I have explored the potential for using the C IV forest down to red-shift z = . a forest. The need for turning to this weakerprobe of intergalactic structure is driven by the fact that Ly a ab-sorption drops out of the optical window at z ≈
2, and the relativeease with which quasars below z ≈ a emission line were in-vestigated between Ly a and C IV in emission, split by the Si IV emission line. Using BOSS DR9 quasar spectra, the metal power,metal populations, and strength of the C IV absorption signal in re-lation to Ly a absorption were tested. I find that C IV absorptiondominates these bands, and represents 80% of absorption. Further-more, the signal associated with the C IV forest at low redshift isa factor of 4–25 weaker that that associated with the Ly a forest athigh redshift, but with greater bias, and a dynamic range suited toprobing higher density regions.The upcoming eBOSS survey will provide five times the ef-fective number density of quasars available for a z < IV forestsurvey compared to the BOSS Ly a forest survey. Therefore it mayprovide broadly comparable accuracy of around 2% in the BAOscale. In combination with the eBOSS quasars redshift distribu-tion, the C IV - quasar cross-correlation may provide uncertainty aslow as 1%. Forthcoming quasar surveys as part of Dark EnergySpectroscopic Instrument (DESI) and WHT Enhanced Area Veloc-ity Explorer (WEAVE) also have the potential to provide improvedconstraints. ACKNOWLEDGMENTS
REFERENCES
Ahn C. P. et al., 2012, ApJS, 203, 21Anderson L. et al., 2014, MNRAS, 439, 83Blake C. et al., 2011, MNRAS, 415, 2892Busca N. G. et al., 2013, A&A, 552, A96Delubac T. et al., 2014, arXiv:1404.1801Font-Ribera A. et al., 2014a, J. Cosmology Astropart. Phys., 5, 27Font-Ribera A., McDonald P., Miralda-Escud´e J., 2012, J. Cos-mology Astropart. Phys., 1, 1Font-Ribera A., McDonald P., Mostek N., Reid B. A., Seo H.-J.,Slosar A., 2014b, J. Cosmology Astropart. Phys., 5, 23 c (cid:13) x RAS, MNRAS , 1–6 Pieri M. M. et al.
Irˇsiˇc V. et al., 2013, J. Cosmology Astropart. Phys., 9, 16Lee K.-G. et al., 2013, AJ, 145, 69McDonald P., Eisenstein D. J., 2007, Phys. Rev. D, 76, 063009McQuinn M., White M., 2011, MNRAS, 415, 2257Meiksin A. A., 2009, Reviews of Modern Physics, 81, 1405Palanque-Delabrouille N. et al., 2013, A&A, 559, A85Pˆaris I. et al., 2012, A&A, 548, A66Pieri M. M., Frank S., Mathur S., Weinberg D. H., York D. G.,Oppenheimer B. D., 2010a, ApJ, 716, 1084Pieri M. M., Frank S., Weinberg D. H., Mathur S., York D. G.,2010b, ApJ, 724, L69Pieri M. M., Haehnelt M. G., 2004, MNRAS, 347, 985Pieri M. M. et al., 2014, MNRAS, 441, 1718Pieri M. M., Schaye J., Aguirre A., 2006, ApJ, 638, 45Rahmati A., Schaye J., 2014, MNRAS, 438, 529Schaye J., Aguirre A., Kim T.-S., Theuns T., Rauch M., SargentW. L. W., 2003, ApJ, 596, 768Slosar A. et al., 2011, J. Cosmology Astropart. Phys., 9, 1Slosar A. et al., 2013, J. Cosmology Astropart. Phys., 4, 26Turner M. L., Schaye J., Steidel C. C., Rudie G. C., Strom A. L.,2014, arXiv:1403.0942Weinberg D. H., Mortonson M. J., Eisenstein D. J., Hirata C.,Riess A. G., Rozo E., 2013, Phys. Rep., 530, 87 c (cid:13) x RAS, MNRAS000