The Population of Weak Mg II Absorbers I. A Survey of 26 QSO HIRES/Keck Spectra1

51 Pages·2009·1.22 MB·English

The Population of Weak Mg II Absorbers I.
A Survey of 26 QSO HIRES/Keck Spectra1
Christopher W. Churchill2, Jane R. Rigby, Jane C. Charlton3
Department of Astronomy and Astrophysics
The Pennsylvania State University, University Park PA 16802
[email protected], [email protected], [email protected]
8
9
9
1 Steven S. Vogt2,4
l Board of Astronomy and Astrophysics
u
University of California, Santa Cruz CA 95064
J
[email protected]
4
1

1
v
1
ABSTRACT
3
1 We present a searchfor “weak” Mgii absorbers [those with W (2796)≤0.3 ˚A] in the HIRES/Keck
7 r
spectra of 26 QSOs. We found 30, of which 23 are newly discovered. The spectra are 80% complete
0
8 to Wr(2796) = 0.02 ˚A and have a cumulative redshift path of ∼ 17.2 for the redshift range 0.4 ≤ z ≤
9 1.4. The number of absorbers per unit redshift, dN/dz, is seen to increase as the equivalent width
/ thresholdis decreased;we obtained dN/dz =1.74±0.10 for our 0.02≤W (2796)≤0.3 ˚A sample. The
h r
p equivalent width distribution follows a power law, N(W) ∝ W−δ, with δ ∼ 1.0; there is no turnover
- down to W (2796) = 0.02 ˚A at hzi = 0.9. Thus, weak absorbers comprise at least 65% of the total
r
o
Mgii absorption population, which outnumbers Lyman limit systems (LLS) by a factor of 3.8±1.1;
r
t the majority of weak Mgii absorbers must arise in sub–LLS environments. Tentatively, we predict
s
a that ∼ 5% of the Lyα forest clouds with Wr(Lyα) ≥ 0.1 ˚A will have detectable Mgii absorption to
v: Wrmin(2796)=0.02 ˚A and that this is primarily a high–metallicity selection effect ([Z/Z⊙]≥−1). The
i individual absorbing “clouds” have Wr(2796)≤ 0.2 ˚A and their narrow (often unresolved) line widths
X
implytemperaturesof∼25,000K.ThesefactsimplythatMgiiabsorbingstructuresfigureprominently
r as tracers of sub–LLS environments where gas has been processed by stars. We compare the number
a
density of Mgii systems with that of both high and low surface brightness galaxies and examine the
range of sizes inferred for Mgii absorbers with W (2796)≥ 0.02 ˚A. We discuss evidence for a range of
r
ionization conditions (low, high, and multi–phase) in the population of systems selected by weak Mgii
absorption and speculate as to the expected nature and redshift evolution of this population.
Subject headings: galaxies: interstellar medium — galaxies: evolution — quasars: absorption lines
Submitted to ApJ
1Basedinpartonobservations obtained atthe W.M.KeckObservatory, whichisjointlyoperated bytheUniversityofCaliforniaand
theCaliforniaInstituteofTechnology.
2VisitingAstronomerattheW.M.KeckObservatory
3CenterforGravitational PhysicsandGeometry,PennsylvaniaStateUniversity
4UCO/LickObservatories,UniversityofCalifornia
1. Introduction sorption. Churchill & Le Brun (1998) have discussed
thispossibilitybaseduponthediscoveryoftwonear–
Wepresentasurveyfor“weak”Mgiiλλ2976,2803
solarto super–solarmetallicity Mgiiabsorbersinthe
absorption in the spectra of QSOs obtained with
Lyα forest of PKS 0454+039. As such, our current
the HIRES spectrograph (Vogt et al. 1994) on the
picture of the relationship between Mgii absorbing
Keck I telescope. There have been several Mgii sur-
gas and star forming environments may require some
veys over the previous decade (Lanzetta, Turnshek, modification. Perhaps the observed L ∼ 0.06L∗
K K
& Wolfe 1987, hereafter LTW; Tytler et al. 1987,
cutoffintheluminosityfunctionofMgiiabsorption–
hereafter TBSYK; Caulet 1989; Petitjean & Berg- selected galaxies or the cut off at ∼ 40h−1 kpc of
eron 1990, hereafter PB90; Sargent, Steidel, & Bok-
Mgii absorption around these galaxies does not ap-
senberg 1988,hereafter SSB; Steidel & Sargent 1992, ply for W (2796) ≤ 0.3 ˚A. Alternatively perhaps an-
r
hereafter SS92). These surveys were complete to a
other population of star forming objects that pref-
rest–frame Mgii λ2796 equivalent width, W (2796),
r erentially give rise to weaker Mgii absorption is im-
of0.3˚Aandabove. Themorecomprehensiveworkof
plied. It may (or may not) be that this hypothetical
SS92 yielded solid statistics on the equivalent width
alternative population exhibits stochastic cosmolog-
distribution, redshift number density, large scale ve-
ical evolution, that the objects do not steadily form
locityclustering,andevolutionovertheredshiftrange
starsormaintainstableabsorbinggasstructures. Or,
0.3≤z ≤2.2, for which the Mgii doublet can be ob-
perhapsbothaslightmodificationtothe currentpic-
served with ground based telescopes.
tureofMgiiabsorbinggalaxiesand theincorporation
The shape of the Wr(2796) distribution function of another population of objects would be implied.
at smaller equivalent widths has important implica- Measuringthe statisticalabsorptionpropertiesofthe
tions for our understanding of cosmic chemical evo- weakestMgiiabsorbersisafirststeptowardverifying
lution and its connection to star producing environ- or casting new light on such speculations.
ments. At 0.3 ≤ z ≤ 1.0, Mgii absorption with Below W (2796)≤0.3 ˚A, BP90 and SS92 inferred
r
W (2796)≥0.3˚AMgiihasbeenfoundtoarisewithin
r a cut off in the Mgii equivalent width distribution,
∼40h−1 kpc5 of normal galaxies (Bergeron& Boiss´e
the number of absorbers per unit redshift per unit
1991; Steidel, Dickinson, & Persson 1994). These
equivalent width. However,the completeness of their
galaxiesexhibit awide rangeofcolorsfromlate–type data below 0.3 ˚A dropped rapidly. Their conclusions
spirals to the reddest ellipticals (though more lumi-
have necessarily been based upon a comparison of
nousgalaxiesareredder),andhaveL andL lumi-
B K the number of Mgii absorbers actually detected with
nosity functions consistent with the local luminosity W (2796)≤0.3 ˚A, correctedfor completeness, to the
r
function(typeslaterthanSdareabsent). Ithasbeen
number of these absorbers predicted by extrapolat-
concluded that a wide range of morphological types
ing the equivalent width distribution. Measurements
are contributing to the Mgii absorbing gas cross sec-
of the equivalent width distribution, however, were
tion, except that isolated low–mass (L ≤ 0.06L∗ )
K K somewhat inconclusive; the data were adequately de-
“faint blue galaxies” are not (Steidel, Dickinson, &
scribed either by a power law distribution with slope
Persson1994). Arapidcutoffintheequivalentwidth
δ = −2.0 (TBSYK) or δ = −1.6 (SS92), or by an
distributionwouldimplythatthisobservedsampleof
exponential with a characteristic equivalent width of
galaxiesprovidesacomplete picture ofthe starform- W (2796) = 0.66 ˚A (SS92; LTW). Womble (1995)
r
ingenvironmentsthatgiverisetoMgiiabsorbinggas.
andTripp, Lu,&Savage(1997)havetentativelycon-
Currently,thatpictureisoneinwhichthegalaxypop-
cluded that the equivalent width distribution contin-
ulationselectedbyMgiiabsorptionisstable,showing ues to rise below 0.3 ˚A. The HIRES/Keck spectra
very little cosmologicalevolution from z ∼1.
obtained for the thesis of Churchill (1997a; also see
If, on the other hand, the number of Mgii ab- Churchill, Vogt, & Charlton 1998) provide duplicate
sorbersperunitredshiftcontinuestorisefordecreas- redshift coverage of 26 of the 103 QSO sight lines
ing Wr(2796), it would be implied that surveys com- studied by SS92. The spectra are 80% complete to
pleteto0.3˚Ahaveunveiledonlyasmallportionofthe Wmin(2796) = 0.02 ˚A, and provide an opportunity
r
populationofmetal–linesystemsselectedbyMgiiab- to directly investigate if in fact there is a paucity of
Mgii absorbers with W (2796) ≤ 0.3 ˚A and to mea-
r
5Throughout this paper we use q0 =0.05 and express physical suretheshapeofthedistributionofequivalentwidths
sizesintermsofh=H0/100.
2
well below 0.3 ˚A. whichstrongsystemsarepresent,inwhichcasethese
Using a database of high quality high resolution QSOspectrawouldnotcompriseanunbiasedsample.
spectra, we have set out to measure the statistical We return to this point in §4.3.
absorption properties of the weakest Mgii systems. Though the sample of spectra are well suited for
The acquisition of the data and their reduction and an unbiasedsurvey, they are not ideally suited. Each
analysis are described in §2. In §3, we present our spectrum was obtained using a single setting of the
adoptedsampleofweakMgiisystems. In§4wecom- HIRESechelleandcrossdisperser. Eachhasbreaksin
putetheredshiftpathdensity,determinetheshapeof wavelengthcoverageredwardof5100˚Abecauseasin-
the equivalent width distribution, examine the clus- glesettingofthe2048X2048CCDcannotcapturethe
tering, and discuss the general absorption properties complete free spectral range beyond this wavelength.
of weak Mgii absorbers. We discuss the ionization The gaps in redshift coverage begin at z ∼ 0.83 and
conditions and some inferences about the luminous becomemorepronouncedtowardhigherredshifts(see
objects associated with these absorber in §5. Our Figure 1a–c). As discussed in §2.4, we searched the
main results are summarized in §6. spectra for the presence of both members of the dou-
blet; bothmembershadtobepresentinorderforthe
2. Observations and Data Analysis system to be included in the survey.
2.1. The Sample of QSO Spectra 2.2. Data Reduction and Analysis
Afulldescriptionofthedataacquisitionisgivenin The raw data frames were bias corrected, flat
Churchill(1997a)andinChurchill,Vogt,& Charlton fielded, cosmic ray cleaned, combined, and scattered
(1998, hereafter CVC98). In short, 26 QSO spectra light corrected, using the standard IRAF7 packages
were obtained with the HIRES spectrometer (Vogt followingthetechniquesdescribedinChurchill(1995)
et al. 1994) on the Keck I telescope. The QSOs and and CVC98. The individual spectra were extracted
their emission redshifts, the UT date of the observa- using the optimal extraction algorithms provided in
tions, the summed exposure times, and the approxi- thenoao.twodspec.apextract package. Thewavelength
mate wavelength coverageare presented in Table 1. calibration was done interactively using the eciden-
These spectra were originally part of a targeted tify task of the noao.imred.echelle package. We have
survey to study the kinematics of known Mgii ab- not resampledthe data by linearizing the wavelength
sorberscompiledby SS92. Thus,eachQSOsightline as a function of pixel along the dispersion direction.
was carefully selected based upon the known pres- Linearizing the data involves a flux conserving inter-
ence of one or more Mgii absorbers. In fact, sight polation between adjacent pixels. This interpolation
lines with the largest equivalent width systems were introduces correlatederrors that are difficult to treat
preferentially selected. These “strong” systems were andmodifiessubtleshapesinabsorptionprofileswhen
expected to reveal rich kinematic structures because the signal–to–noise ratio is high.
of the observed correlation of the number of compo- Each full spectrum was then converted into an
nents with equivalent width (PB90). As such, the ASCII file andthe remainderofthe analysiswasper-
spectra comprise a biased sample of QSO sight lines formed using our own software (further details are
with “strong” Mgii absorbers. presentedinChurchill1997a). Becauseofthe HIRES
The SS92 survey had a fairly uniform Mgii λ2796 echelle format and the large size of the CCD, there
rest–frameequivalent width limit of 0.3 ˚A, where the is redundant wavelengthcoverage(at the orderends)
5σlimithasbeenquoted. Therefore,thelinesofsight below 5100˚A. There are welldocumented techniques
presented by CVC98 are actually unbiased for Mgii available in IRAF for combining these regions using
systemsthathaveW (2796)≤0.3˚A,since nothingis interpolation to resample and combine the data into
r
known a priori about the presence of these “weak” a “single” spectrum. We have chosen to not resam-
Mgii systems6. However, it is possible that weak ple and combine the regions of redundant coverage.
systems are preferentially found in sight lines along Thecontinuumfittingisacriticalstepintheanalysis
6From this point forward, we refer to “weak” systems as those 7IRAF is distributed by the National Optical Astronomy Ob-
having Wr(2796) ≤0.3 ˚A and call Wr(2796) >0.3 ˚A systems servatories,whichareoperatedbyAURA,Inc.,undercontract
“strong”. totheNSF.
3
because it results in correlated errors between pix- itylimitsfordetectingaλ2796transitionasafunction
els. A searchfor weak features could easily be biased of redshift, are shown in Figure 1a–1c. Where there
by systematics in the continuum fit. Thus, we took is no redshift coverage (interorder gaps), we have ar-
great care with the fitting. The adopted continuum bitrarily set the limiting equivalent width to zero.
fits, based upon the formalism of Sembach & Savage
(1992), were obtained by minimizing χ2 between the 2.3. NotesontheReductionofSelectedQSOs
fluxvaluesandasmoothfunction,usuallyaLegendre
2.3.1. Q0002+051; z =1.899
em
polynomial.
During the extraction, the uncertainty in the flux Only a single HIRES spectrum on this QSO was
ineachpixelwascomputedbaseduponasimplevari- obtained so that removing cosmic rays, especially
ancemodelthataccountedforPoissonnoiseandread from the sky, was very problematic. As a conse-
noise. The uncertainty spectra are critical, as we quence, the zero level of the spectrum is uncertain
shall discuss below, for the objective identification by ∼ 10% and the measured equivalent widths may
of absorption features, especially those that are un- be biased by this probable zero–point offset, which
resolved. If the values in the uncertainty spectrum varied from echelle order to echelle order. This un-
are smaller than the noise level of the nearby contin- certainty has not been included in the error estimate
uum, then too many absorption (and emission) fea- of the quoted equivalent widths.
tures would be objectively identified to the desired
2.3.2. Q1548+092; z =2.749
significance level; if the uncertainty spectrum values em
are larger than the noise in the data, then weak but The region of the QSO spectrum we observed is
real absorption features would not be found. blueward of the Civ emission line by ∼ 100 ˚A. The
While the resulting uncertainty spectra accurately Lyα forest compromises all transitions of metal line
depicted the relative pixel to pixel uncertainties, we systemsthatliebluewardof∼4560˚A.Thus,wehave
foundthattheirvalueswereoftennotagooddescrip- notsearchedforMgiidoubletsinthespectrumbelow
tion of the noise level in the continuum. Thus, for thiswavelength;wecouldnotconfidentlydetectMgii
each echelle order, we enforced a unity χ2 between systems below z ∼ 0.62. Observations of this QSO
ν
thesmoothfittedcontinuummodelandthespectrum were plagued by 1–2 magnitudes of extinction due to
byscalingtheuncertaintyspectrumbyasingle“opti- cirruscloudcover. Assuch,thesignal–to–noiseratio,
mizing”multiplicativefactor,f. Toobtainf,weroot and thus detection sensitivity, is very low.
solvedthe function 1−χ2, where χ2 =V /fV , and
ν ν fit σ
V and V are the variance in the continuum fit and 2.4. Doublet Searching
fit σ
the uncertainty spectrum, respectively. As a result,
We have searched for redshifted Mgii doublets in
the uncertainty spectrum retained its overall shape
our 26 QSO spectra. The sensitivity of the search
andpixeltopixelstructure,butthepixelvaluesmore
is not uniform. This is due to the varying observing
accurately reflected the noise level of the continuum
conditions,therangeinthebrightnessoftheobserved
flux values. The process requiredaniterative conver-
QSOs, and the wavelength (redshift) dependent sen-
gence algorithm because it was critical to objectively
sitivity of the HIRES spectrograph. We will return
mask absorption features that would otherwise bias
to the issue of non–uniform sensitivity in §3.3 and in
the optimizing scaling factor (only continuum pixels
§4.
were used for the root solving in each iteration).
The search for Mgii doublets involved the follow-
For the objective identification of unresolved ab-
ing steps. First, a complete list of features were ob-
sorption (and emission) features, we have used the
jectively defined in each QSO spectrum. All features
formalism presented by Schneider et al. (1993), who
had a 5σ detection significance level. To locate can-
developedatechnique tocomputebothanequivalent
didateMgiidoublets,wethentestedthefeaturesone
width spectrum and an equivalent width uncertainty
by one, starting at the smallest wavelength feature
spectrumbaseduponoptimalpixelbypixelweighting
and moving toward larger wavelengths. We assumed
with the instrumental spread function. The equiva-
each feature was a candidate λ2796 line with ob-
lentwidthuncertaintyspectraprovidetherest–frame
served central wavelength λ . Then, the expected
limitingequivalentwidthoftheMgiiλ2796transition 27
location of the λ2803 feature was computed from
as a function of redshift. The 5σ rest–frame sensitiv-
4
.1
.05
0
.1
.05
0
.1
.05
0
.1
.05
0
.4 .6 .8 1 1.2 1.4 1.6 1.8 .4 .6 .8 1 1.2 1.4 1.6 1.8 .4 .6 .8 1 1.2 1.4 1.6 1.8
Fig. 1a.— The survey sensitivity, given as the 5σ Wmin(2796), is shown as a function of redshift. Breaks in the
r
echelle spectral coverage, where there is no possibility of detecting a Mgii doublet in a given redshift range, are
arbitrarily assigned Wmin(2796)=0.
r
λ = 2803.531(λ /2796.352). Centered about λ , tal redshift window of ∆z ∼ 0.02, or ±3000 km s−1
28 27 28
an equivalent width and its uncertainty were mea- aboutthefeature. Thefalsealarmprobabilityissim-
sured in an aperture with the same full width at the plythefractionofpixelswithdetectedfeatures(both
continuum as that of the candidate λ2796 feature. emission and absorption) having a significance level
These quantities were measured using the formalism greaterthan orequalto the candidateλ2803feature.
ofSembach&Savage(1992). Thepairwasdesignated
as a candidate doublet when the λ2803detection sig- 2.5. Defining the Doublet Sample
nificance was roughly equal to or greater than half
The criteria by which a candidate Mgii doublet
that of the λ2796 feature (given by the ratio of the
is determined is described above. In practice, the
fλ) such that the doublet ratio was consistent with
detection levels of the λ2803 transitions for all in-
1≤DR≤2 within the 1σ uncertainties.
cluded doublets were greater then 4.5σ. Often, other
We also employed a quantitative measure of the species, particularly strong Feii transitions and the
chance that the candidate λ2803 transition could be
Mgi λ2853 transition, were detected. We did not
a random feature or a transition from another sys- consider the detection of Feii or Mgi as part of the
tem8 alongthe QSOsightline. A“falsealarm”prob-
selection criteria for our adopted sample.
abilityiscomputedbyscanningthespectrumpixelby
MostbonafideMgiidoubletshavefalsealarmprob-
pixel(alsowithanaperturegivenbythefullwidthat abilities of P ≤ 10−6. We found nine candidate
fa
the continuum of the candidate λ2796 feature). The
Mgii doublets with 10−6 ≤ P ≤ 0.05. Most of
scan is performed about 50 ˚A to both sides of the fa
these had unphysical doublet ratios or their wave-
candidate λ2803 feature. This corresponds to a to-
lengths did notalignwithin 3σ oftheir uncertainties;
thus they were removed from the candidate list. The
8Wehaveperformedanexhaustivecross–checkingoflineidenti-
largest false alarm probability for a Mgii doublet in
ficationsfromCivandMgiiredshiftsreportedintheliterature
and from those discovered in our spectra. Details of the line our adopted sample is Pfa ∼ 0.009 (see discussion of
identificationsandthelistofsearchedtransitionsandredshifts S24 in Q1213−003atz =1.1277). Visual inspection
willbepresentedinCVC98.
of this system reveals it to be robust; the relatively
5
.1
.05
0
.1
.05
0
.1
.05
0
.1
.05
0
.4 .6 .8 1 1.2 1.4 1.6 1.8 .4 .6 .8 1 1.2 1.4 1.6 1.8 .4 .6 .8 1 1.2 1.4 1.6 1.8
Fig. 1b.— Same as for figure 1a
large P is due to the number of features of equal andtheMgiidoubletratio. Thelastcolumncontains
fa
or greater significance within the 6000 km s−1 win- thecumulativeredshiftpath,Z(W ,DR),whichisthe
r
dow. Taken at face value, this implies that the S24 total redshift path over which the tabulated system
absorberhada0.9%chanceofbeingafalsedetection. could have been detected in this survey (see Eq. 2).
We are quite certain that our adopted sample is not The velocity widths, ω , are measured directly from
v
contaminated by false Mgii systems. the flux values according to
In addition to the above constraints, a nearest
v+ v+
neighborvelocityseparationofgreaterthan500km s−1 ω2 = τ (v)(∆v)2dv τ (v)dv, (1)
v Z a (cid:30)Z a
was enforced. Of the strong Mgii absorbing sys- v− v−
temsstudiedbyCVC98,twohadamaximumvelocity
where τ (v) = ln[I (v)/I(v)] is the apparent optical
spread slightly greater than 450 km s−1, if they are a c
depth (Savage & Sembach 1991), I (v) is the contin-
in fact single systems. The velocity filter was applied c
uumfluxatvelocityv,I(v)isthemeasuredfluxatv,
so that small equivalent width, high velocity compo-
∆v = v−hvi, and hvi is the velocity centroid of the
nents in strong systems would not be included in the
sample. In practice, the 500 km s−1 separationcrite- absorptionprofile. Theωv aremathematically equiv-
alentto the Gaussianwidth of a normaldistribution.
rion resulted in our dropping only a single potential
In Table 3, we present the properties of each sys-
weak Mgii absorber from our sample (see discussion
of S19 in the Q1206+459 spectrum). Had we ap- tem, including the transition identity, the observed
plied a minimum velocity separation of 800 km s−1, wavelength, and the rest–frame equivalent width or
its 3σ upper limit. The equivalent widths and their
we would have obtained the same sample.
uncertaintiesarecomputedusingthemethodsofSem-
3. The Doublet Sample bach & Savage (1992). Because the Feii and Mgi
λ2853 transitions are the strongest and most com-
TheadoptedsampleispresentedinTable2. Tabu- monly found in Mgii absorbers, we have presented
latedarethesystemnumber,the absorptionredshift, their limits for the individual systems. Other transi-
the QSO spectrum in which it was detected, the ve- tions are included when they have been detected at
locity width in km s−1 of the λ2796 transition, the the 3σ level or above.
rest–frame equivalent width of the λ2796 transition,
6
conductedasemi–rigoroussearchfortheCivdoublet
in the FOS/HST spectra of Dobrzycki et al. (1998).
The 3σ detection threshold of these spectra ranged
.1 fromWrmin(1548)=0.15˚Ato0.3˚A.Wealsosearched
the literature for reported Civ in FOS/HST spectra.
We found Civ in six of the fifteen systems for which
.05
FOS/HST spectra were available.
In total, Civ was found or reported in nine of the
0
18 weak Mgii systems for which information on Civ
.1
was available. Thus, it is reasonable to suggest that
roughlyhalfofallweakMgiiselectedabsorberswould
.05
show Civ absorption to the equivalent width thresh-
olds available in existing data. We defer a discussion
0 of the implications until §5.2.
.4 .6 .8 1 1.2 1.4 1.6 1.8
3.2. Discussion of Systems
The spectra for each of the detected transitions in
Fig. 1c.— Same as for figure 1a
the adopted Mgii sample are shown in Figure 2a–
dd. Only those formally detected with a greaterthan
3σ level are shown. Features marked with a “∗” are
3.1. Detection of Other Species
either members of other systems or are unidentified.
3.1.1. In the HIRES Spectra Inacompanionpaper(Churchilletal.1998,hereafter
Paper II), we will examine the detailed absorption
When a candidate Mgii doubletwas identified, we
properties on a system by system basis.
searched for the presence of other transitions in ab-
sorption. Our search included the transitions pre-
3.2.1. S1 (Q1421+331; z =0.45642)
abs
sentedinTable4(describedin§4.4). Sincethesearch
was based upon a priori knowledge of the Mgii ab- There is no previous report of S1 in the litera-
sorption redshift, the detection significance level was ture. NoFeiitransitionswerecapturedbytheHIRES
relaxed to 3σ. As seen in Figure 1a–c, the sensitiv- CCD, only the Mgii and Mgi λ2853 transitions with
ity is a strong function of redshift in the sense that thelatternotbeingdetected. TheMgiiprofilesreveal
we are less sensitive to lower redshift absorbers. As asinglefeaturethatexhibits someunresolvedsatura-
such, transitions with wavelengths blueward of the tion. There is a slight suggestion of blue asymmetry
Mgii doublet were less likely to be detected, espe- in the λ2803 transition that may be due to multiple
cially in the lower redshift systems. narrowcomponents. However,thesignal–to–noisera-
tioislowenoughthattheprofileshapeisnotwithout
3.1.2. LiteratureSearchandFOS/HSTArchivalSpec- some ambiguity. A FOS/HST spectrum of this QSO
tra was not available for our Civ search. Though this
field has been studied in some detail, to date there is
We have searched for the Civ λλ1548,1550 dou-
nogalaxycandidateassociatedwiththisabsorber(C.
blet9attheredshiftsofourMgiisample. Thepurpose
Steidel, private communication).
of the search was to obtain indications of high ion-
ization in metal–line systems selected by weak Mgii
3.2.2. S2 (Q1329+412; z =0.500786)
abs
absorption. From the literature, we found Civ was
reportedin threeofthe sevenhigherredshift systems S2 was first reported as a “probable” Mgii dou-
in which it could be observed from the ground. The blet by SSB. The signal–to–noise ratio of the HIRES
remaining four of the seven have not been observed. spectrumisfairlylow,andonlytheMgiidoubletwas
With the kind assistance of A. Dobrzycki, we also detected. The limiting equivalent widths of the two
captured Feii transitions are not very stringent. The
9WelimitedoursearchtoCiv. Amoredetailedsearchforother profiles are resolved into a broad multi–component
specieswillbepresentedinPaperII. structure, particularly in the higher quality λ2796
7
placedby interpolatingbetweenits nearestneighbors
and its uncertainty has been proportionately scaled
to reflect the “correction factor”. The Mgii profile
1 appears to be a single component that may be par-
tially resolvedin that there is a red–wing asymmetry
in both transitions of the doublet. Though Feii and
0
Mgi were captured by the CCD, only the Mgii dou-
blet was detected. Civ absorption was not detected
1
in the FOS/HST spectrum (also see Jannuzi et al.
1998). Thoughthis QSO field has about five galaxies
0 allwithin±200km s−1 ofz =0.45,whicharepartof
aforegroundcluster,todatethereisnogalaxycandi-
-200 -100 0 100 200
date associated with this weak absorber (C. Steidel,
private communication).
Fig. 2a.— The detected transitions ofthe weak Mgii
systems. All detections are 5σ unless annotated oth-
erwise. Absorption features marked with “∗” are not
1
associated with the system being presented. In sys-
tems with mutliple components, ticks mark the com-
ponents that are detected at the 3σ level.
0
transition. Mgi wasnotdetected. A FOS/HSTspec- 1
trumofthisQSOwasnotavailableforourCivsearch.
There is an associated galaxy with this absorber (C.
0
Steidel, private communication).
-200 -100 0 100 200
Fig. 2c.— Same as for figure 2a
1
3.2.4. S4 (Q1222+228; z =0.550202)
0 abs
There is no previous report of S4 in the literature.
1 The Mgii profile is well resolved with velocity dis-
persion of ω ∼ 10 km s−1, and is likely comprised
v
of multiple narrower components. It is fairly broad
0
and shallow, such that the λ2803 transition equiva-
-200 -100 0 100 200 lent width could be sensitive to the continuum fit;
it is possible the continuum fit could result in the
doublet search algorithm missing or falsely detect-
ing this system. Thus, the continuum fit was exam-
Fig. 2b.— Same as for figure 2a
ined. Both a fit to the full echelle order and a more
constrained fit to the region ±500 km s−1 about the
3.2.3. S3 (Q1354+193; zabs =0.52149) transitionyielded Wr(2803)measurements consistent
within 1σ errors. For both fits, the λ2803 transition
There is no previous report of S3 in the literature. hadagreaterthan5σ detectionandthe doubletfalse
The Mgiiλ2796transitionwascoincidentwithabad alarm ratio was ≤ 10−6. Though Feii and Mgi were
column in the HIRES CCD. This pixel has been re- capturedby the CCD, only the Mgiidoublet wasde-
8
tected. However, the Feii λ2600 transition may have
been detected at the 2.7σ level. Civ absorption was
notdetectedinthe FOS/HST spectrum(alsoseeIm-
pey et al. 1996). Though this field has been studied 1
in some detail, to date there is no galaxy candidate
associatedwiththisabsorber(C.Steidel,privatecom-
0
munication).
1
0
1
-200 -100 0 100 200
0
Fig. 2e.— Same as for figure 2a
1
be biased by a zero–point uncertainty, which has not
0
beenincludedintheerrormeasurement. Civabsorp-
-200 -100 0 100 200 tionwasnotdetectedintheFOS/HSTspectrum(also
see Jannuzi et al. 1998).
Fig. 2d.— Same as for figure 2a
3.2.5. S5 (Q1241+174; z =0.55844) 1
abs
There is no previous report of S5 in the literature.
The Mgii profile is also well resolved with velocity 0
dispersion of ω ∼17 km s−1. Though Feii and Mgi
v
werecapturedbytheCCD,onlytheMgiidoubletwas 1
detected. Associated Civ absorption was tentatively
detected in the FOS/HST spectrum from our search.
0
However, the λ1548 line may be Lyα (Jannuzi et al.
1998). Thoughthisfieldhasbeenstudiedinsomede- -200 -100 0 100 200
tail, to date there is no galaxy candidate associated
with this absorber (C. Steidel, private communica-
tion).
Fig. 2f.— Same as for figure 2a
3.2.6. S6 (Q0002+051; z =0.59149)
abs
S6 was discovered by C. Steidel and collaborators 3.2.7. S7 (Q0454+039; zabs =0.64283)
(C. Steidel, private communication) and is known to
S7 was first reported by Churchill & Le Brun
beassociatedwithared≃1.3L∗ galaxywithimpact
K (1998), who found it to have near–solar metallicity
parameter ≃ 24h−1 kpc (Churchill, Steidel, & Vogt
based upon photoionization modeling of Mgii, Feii
1996). The Mgii profiles are partially resolved, at
and the Lyα transition detected in the FOS/HST
best, and are probably unresolved. Though multiple
spectrum of Boiss´eet al. (1997). Civ absorptionwas
FeiitransitionswerecapturedbytheCCD,nonewere
not detected in the FOS/HST spectrum. Interest-
detected. If Mgi λ2853 is present, it was detected
ingly, there is no galaxy at this redshift in this well
onlyatthe2σsignificancelevel. Asmentionedabove,
studied field (Churchill & Le Brun 1998). The Mgii
theequivalentwidthsmeasuredinthisspectrummay
9
profiles,both symmetric with little structure, are ap-
parentlyunresolved;thereisslightunresolvedsatura-
tion in the λ2796 profile. Here, we also report a 3.3σ
detection of Feii λ2587. Mgi was not detected. 1
0
1 1
0 0
-200 -100 0 100 200
1
0 Fig. 2h.— Same as for figure 2a
1
0
1
1
?
0
0
1
1
0
0
-200 -100 0 100 200 1
0
Fig. 2g.— Same as for figure 2a
1
3.2.8. S8 (Q0823−223; z =0.705472)
abs
0
There is no previous report of S8 in the literature.
TheMgiiprofilesrevealasingleresolvedfeaturewith -200 -100 0 100 200
a small blueward asymmetry. Though Feii and Mgi
werecapturedbytheCCD,onlytheMgiidoubletwas
detected. AFOS/HSTspectrumofthisQSOwasnot Fig. 2i.— Same as for figure 2a
available for our Civ search.
There is a tentative, 3σ detection of Feii λ2600,
3.2.9. S9 (Q0058+019; z =0.72518)
abs
which we present in Table 3, but this detection is
There is no previous report of S9 in the literature. deemed insecure based upon the fact that there are
The Mgii profiles are resolved,with ωv ∼11 km s−1, three3σdetectionsinavelocitywindowof±500km s−1
and saturated in their cores. This system is interest- centered on the putative Feii detection. Moreover,
ing inthat relativelystrongMgi λ2853was detected. thereisnoevidenceto3σforthestrongerλ2383tran-
10