Partial covering of emission regions of Q 0528-250 by intervening H$_2$ clouds

1.5 MB·English

by  V. V. Klimenko

Mon.Not.R.Astron.Soc.000,1–13(2015) Printed29January 2015 (MNLATEXstylefilev2.2)
−
Partial covering of emission regions of Q 0528 250 by
intervening H clouds.
2
V.V. Klimenko1,2, S.A. Balashev1,2, A.V. Ivanchik1,2, C. Ledoux4,
P. Noterdaeme3, P. Petitjean3, R. Srianand5, D.A. Varshalovich1,2
5
1
0
2 1Ioffe Physical-Technical Institute of RAS, Polyteknicheskaya26, 194021 Saint-Petersburg, Russia

n 2St.-Petersburg State Polytechnical University,Polyteknicheskaya29, 195251 Saint-Petersburg, Russia
a
3UniversitéPierre et Marie-Curie, Institut d’Astrophysique de Paris, CNRS-UMR7095, 98bis boulevard Arago, 75014 Paris, France
J
4European Southern Observatory,Alonso de Córdova 3107, Casilla 19001, Vitacura, Santiago 19, Chile
8
2 5 Inter-UniversityCentre for Astronomy and Astrophysics, Post Bag 4, Ganesh Khind, Pune 411 007, India

]
A
G
Accepted 15.12.2014Received08.10.2014
.
h
p
- ABSTRACT
o We present an analysis of the molecular hydrogen absorption system at zabs = 2.811
r in the spectrum of the blazar Q0528−250.We demonstrate that the molecular cloud
t
s does not cover the background source completely. The partial coverage reveals itself
a
[ as a residual flux in the bottom of saturated H2 absorption lines. This amounts to
about (2.22±0.54)% of the continuum and does not depend on the wavelength.This

1 valueissmallanditexplainswhythiseffecthasnotbeendetectedinpreviousstudies
v ofthisquasarspectrum.However,itisrobustlydetectedandsignificantlyhigherthan
0 the zero flux level in the bottom of saturated lines of the Lyα forest, (−0.21±0.22)
1
per cent. The presence of the residual flux could be caused by unresolved quasar
2
7 multicomponents, by light scatteredby dust, and/orby jet-cloud interaction. The H2
absorptionsystemisverywelldescribedbyatwo-componentmodelwithoutinclusion
0
. ofadditionalcomponentswhenwetakepartialcoverageintoaccount.Thederivedtotal
1 column densities in the H2 absorption components A and B are logN(H2)[cm−2] =
0 18.10±0.02and17.82±0.02,respectively.HDmoleculesarepresentonlyincomponent
15 B. Given the column density, logN(HD)=13.33 ± 0.02, we find N(HD)/2N(H2) =
(1.48±0.10)×10−5,significantlylowerthanpreviousestimations.Wearguethatitis
:
v crucial to take into account the partial coverage effects in any analysis of H2 bearing
Xi absorption systems, in particular when studying the physical state of high-redshift
interstellar medium.
r
a Key words: cosmology:observations,ISM:clouds, quasar:individual:Q0528−250
1 INTRODUCTION Direct imaging ofthespatial structureofAGNsispos-
siblewithcurrentinstrumentsmainlyprobinglongerscales.
Todate,severalinterferometricstudiesofthecentralengine
As a result of their cosmological distances quasars
ofthebrightestAGNs(e.g.Jaffe et al.2004;Tristram et al.
(QSOs) appear as point-like objects. Various studies have
2014; López-Gonzaga et al. 2014) have revealed the exis-
aimed to explore the detailed inner structure of quasars,
tenceofahot,parsec-scalediskthatissurroundedbywarm
which isunresolvedevenfor low-redshift activegalactic nu-
dust extended in the polar direction. In the optical band
clei (AGNs), because of their remote distances and sub-
the geometry of the emission line region is investigated by
parsecscalesoftheiremissionregions.InthestandardAGN
indirectmethods.Reverberationmappingestablishesthere-
paradigmthecentralregionisdividedintoanaccretiondisk,
lationship between the size and the luminosity of the BLR
adusty-torus,ajet,abroadlineregion(BLR),andanarrow
andyieldsatypicalBLRsizeofR ∼0.2pc(Kaspi et al.
line region (NLR). Each of these regions contribute differ- BLR
2007; Chelouche & Daniel 2012) for high redshift luminous
ently to theAGN emission spectrum.
(cid:13)c 2015RAS
2 Klimenko et al.
quasars. Differential microlensing allows for a constraint on tial coverage for the H cloud at z = 2.811 in the spec-
2 abs
the accretion disk size <∼ 3×10−3pc (Blackburneet al. trum of Q0528−2508 is presented in this study. We anal-
2011; Jiménez-Vicente et al. 2012) and for an estimation of yse a new spectrum and detect residual flux in the bottom
the size of the BLR ∼ 0.1 pc (Sluse et al. 2011). The ob- of saturated H lines (J=0 and J=1 levels). In case this
2
servationsofgamma-ray emission constrain thesizeofajet flux is not taken into account, saturated lines yield large
constituent to a few parsecs (Abdoet al. 2010). χ2 valuesandamulticomponentmodelisusedinstead(e.g.
Another estimate of the size of the AGN emitting re- Kinget al. 2011).
gions comes from constraints derived from covering factor The remainder of this paper is organized as follows. A
analysis of intervening H bearing clouds which happen to brief description of thedata is given in Section2. The prin-
2
cover the background source only partially. Analysis of the ciples of partial coverage are described in Section3. In Sec-
partialcoverageofQ1232+082byamolecularhydrogenab- tion4,we presenttheanalysis of theH2 absorption system,
sorptioncloudallowedBalashev et al.(2011)toestimatethe accountingforpartialcoverage.TheHDmolecularlinesare
size of theCIV BLR, RCIV ∼ 0.16+−00..0181pc. exploredinSection5.TheresultsarediscussedinSection6,
Molecular hydrogen absorption systems, a subset and we give a brief conclusion in Section7.
of damped Lyα systems (DLAs) and sub-damped Lyα
systems (sub-DLAs), reveal diffuse and translucent in-
terstellar clouds in high redshift intervening galaxies
2 DATA
(Noterdaeme et al. 2008). An analysis of H absorption
2
systems allows for examining the physical conditions of The molecular hydrogen was identified for the first
diffuse clouds in distant galaxies (Srianand et al. 2005; time at high redshift in the very spectrum of Q0528−2508
Noterdaeme et al. 2007). It has been shown that the gas (Levshakov& Varshalovich1985).Thisquasarwasobserved
is a part of the cold neutral medium with comparatively many times, in particular during the period between 2001
low kinetic temperature (T ∼ 50−500K) and high den- and 2009 using both spectroscopic arms of Ultraviolet and
sities (nH > 10 cm−3), thus compact sizes (l <∼ 1pc). Visual Echelle Spectrograph (UVES) of the Very Large
Comparison of 21 cm and H2 absorptions suggests that Telescope (VLT); for a description of the instrument, see
the H2 absorption originates from a compact gas that Dekkeret al.(2000).Thelogoftheobservationsusedinour
probably contains only a small fraction of Hi measured work is shown in Table 1. These observations relate to four
along the line of sight (Srianand et al. 2012). These sys- programmes, three of which were carried out in 2001–2002:
tems are important instruments for the analysis of sev- 66.A-0594(A)(PI:Molaro),68.A-0600(A)(PI:Ledoux),and
eral cosmological problems, as follows. (i) The discov- 68.A-0106(A) (PI:Petitjean). Theinstrument settingsused
ery of HD/H2 clouds at high redshift (Varshalovich et al. duringtheseobservationswere a1-arcsec slit and 2x2CCD
2001) provides an independent way to estimate the pri- pixelbinninginbotharms,resultinginaresolvingpowerof
mordial deuterium abundance (D/H) and therefore the R∼45000intheblueandR∼43000inthered.Therewasno
relative baryon density of the Universe Ωb which is one ThArlampcalibrationattachedtoeachoftheexposures.An
of the key cosmological parameters (Balashev et al. 2010; additionalseriesofobservationswasperformedin2008-2009
Ivanchiket al. 2010). (ii) The comparison of H2 wave- underprogramme082.A-0087(A)(PI:Ubachs).Thesettings
lengths observed in QSO spectra with laboratory ones forthatprogrammewerea0.8-arcsecslitinthebluearmand
(justforthisquasarQ0528−250,Varshalovich & Levshakov a0.7-arcsec slit in thered. The2x2CCD pixelbinningwas
(1993), Cowie & Songaila (1995), Potekhin et al. (1998), alsousedatthattime.Thisresultedinaresolvingpowerof
Ubachs& Reinhold (2004), King et al. (2011)) allows us to R∼60000 in the blue and R∼56000 in the red. Because the
test the possible cosmological variation of the proton-to- goal of that programme was to set a limit on the variation
electronmassratioµ=mp/me.Becausethede-composition of µ, ThAr lamp calibrations were also taken immediately
of H2 absorptions into several components is crucial for after each observation.
studies of the fundamental constant variability problem, The data presented in Table 1 were reduced using
we should pay attention to the partial coverage effect. It theUVESCommon PipelineLibrary(CPL) datareduction
is known that taking into account the partial coverage ef- pipelinerelease4.9.5usingtheoptimalextractionmethod1.
fects,thephysicalmodeloftheabsorptionsystemdiffers(see The inter-order background (scattered light inside the in-
Balashev et al.2011).(iii)Theinterpretationoftherelative strument) was carefully subtracted in both the flat-field
populations of Ci fine-structure excitation levels and CO framesandthescienceexposures.Linearsplineinterpolation
rotational levels (Srianand et al. 2000; Noterdaeme et al. was used to produce a two-dimensional background image,
2011)allows ustomeasurethetemperatureTCMB(z)ofthe whichwas subsequentlysmoothed usinganaverage boxcar.
cosmic microwave background radiation at high redshift. Fourth-order polynomials were used to find the dispersion
Here,wearguethatitisnecessarytotakeintoaccount solutions.However,theerrorsonlyreflectthecalibrationer-
the partial coverage of quasar emission regions by a com- ror at the observed wavelengths of the ThAr lines used for
pact intervening H2 cloud in order to derive a robust fit wavelength calibration. All the spectra were corrected for
of the absorption lines. If this effect is not taken into ac- themotion of theobservatory around thebarycentreof the
count, column densities can be underestimated by a factor Solar system and then converted to vacuum wavelengths.
of up to two orders of magnitude. The first case of such These spectrawere interpolated intoacommon wavelength
an analysis has been presented by Balashev et al. 2011 for
Q1232+082. The second case of partial coverage has been
detected by Albornoz Vásquezet al. (2014) for H2 bearing 1 see the UVES pipeline user manual available for download at
cloudtowardsthequasarQ0643−504.Thethirdcaseofpar- ftp://ftp.eso.org/pub/dfs/pipelines/uves/uves-pipeline-manual-22.8.pdf
(cid:13)c 2015RAS,MNRAS000,1–13
Partial covering of Q0528−250 by intervening H clouds. 3
2

Table 1.Logoftheobservations
a) b) c)
No. UTDate Program ID Exposure Slit s
(sec) (arcsec) unit d
12 0034..0022..22000011 6666..AA--00559944((AA)) 12××55665555 11..00 bitrary Ftotal Fclou
r
a
3 05.02.2001 66.A-0594(A) 1×5655 1.0 x,
u
4 07.02.2001 66.A-0594(A) 1×5655 1.0 Fl

5 13.02.2001 66.A-0594(A) 1×5655 1.0 LFR Ftotal
6 13.03.2001 66.A-0594(A) 1×5655 1.0
7 15.03.2001 66.A-0594(A) 1×5655 1.0
Wavelength
8 17.10.2001 68.A-0600(A) 1×3600 1.0
Figure 1.Anillustrationoftheeffectofpartialcoverage onthe
9 18.10.2001 68.A-0600(A) 2×3600 1.0
absorption-line profiles. Different panels show (a) a highly satu-
10 08.01.2002 68.A-0106(A) 2×3600 1.0
ratedlinewithtotalcoverage,fc=1, (b)ahighlysaturatedline
11 09.01.2002 68.A-0106(A) 2×3600 1.0 withpartialcoverage,fc=0.8 and(c)apartiallysaturatedline
12 10.01.2002 68.A-0106(A) 2×3600 1.0 with the same partial coverage, fc = 0.8. The line flux residual
13 23.11.2008 082.A-0087(A) 2×2900 0.8-0.7 (LFR)is thefraction of the QSOflux that isnot intercepted by
the cloud. It can be easily derived from the spectral analysis in
14 25.11.2008 082.A-0087(A) 1×2900 0.8-0.7
case(b),butitisnotstraightforward todetect incase(c).
15 23.12.2008 082.A-0087(A) 4×2900 0.8-0.7
16 25.01.2009 082.A-0087(A) 1×2900 0.8-0.7
17 26.01.2009 082.A-0087(A) 1×2900 0.8-0.7
illustrated in Fig.1. The determination of the covering fac-
18 26.02.2009 082.A-0087(A) 1×2900 0.8-0.7
toristrivialin thecase ofhighlysaturated absorptionlines
Total 108450 (seeFig.1panels(a)and(b))whileforapartiallysaturated
line(seeFig.1case(c))theanalysisrequiresamoresophis-
ticated procedure. In this case it is necessary to use several
absorption lines originating from the same levels but with
arrayandgeneratedtheweighted-meancombinedspectrum differentvalues of λf,which istheproduct of theoscillator
using theinverse squares of errors as weights. All theavail- strength, f, and the wavelength of the transition, λ. Such
able exposures were utilized to increase the signal-to-noise an analysis has been performed by Ivanchiket al. (2010)
ratio up to ∼60 per pixel in the wavelength range of the and more precisely by Balashev et al. (2011) for the spec-
H2 absorption lines (at z = 2.811). As shown below, this trumofQ1232+082,andbyAlbornoz Vásquezet al.(2014)
allows us to detect and study the effect of partial coverage for the spectrum of Q0643−504. A similar situation was
in details. observed for HE0001−2340. Jones et al. (2010) have con-
sidered thepossibility ofpartial coverage oftheBLR toex-
plaintheobservedMgiiequivalentwidths.Incontrasttothe
rare situations where partial coveringoccurs from interven-
3 EFFECT OF PARTIAL COVERING
ingsystems(seealso Petitjean et al.2000),partial covering
Partial covering implies that only a part of the back- is typical for absorption systems associated with quasars (
ground source is covered by the absorbing cloud. Mainly e.g. Petitjean et al. 1994; Rupkeet al. 2005; Hamann et al.
this can bea consequence of theabsorbing cloud size being 2010; Muzahid et al. 2013).
comparableto,orevensmallerthan,theprojectedextentof Afailuretotakeintoaccountthepartialcoverageeffect
thebackground source. inaspectroscopicanalysiscanleadtoasignificantunderes-
Partialcoveringisreadilydetectableinthespectrumof timationofthecolumndensityofanabsorber.Thesystem-
abackgroundquasarifthecoresofthesaturatedabsorption atic bias (of column density) can exceed several orders of
lines do not reach the zero flux level. This indicates that a magnitude for saturated lines. As an example, consider an
partoftheradiationfromtheQSOpassesbythecloud.The absorptionlinewhichconsistsofonecomponentandhasthe
covering factor characterizing partial coverage is definedby high column density. The spectrum and the corresponding
theratio, one-component model are shown in panel (a) of Fig. 2. In
panels (b), (c), (d) and (e) the same line is presented, but
F
fc = Fcloud, (1) part of radiation from background source (10% for clarity)
total
passesbyacloud.IfwetakeintoaccounttheLFR,thenthe
where Fcloud is the flux that passes through the absorbing line can be properly fitted by a one component model and
gas andFtotal isthetotal flux.Therefore themeasured flux we can recover the high input column density (with given
in thespectrum, F(λ),can be written as: accuracy) and measure theLFR value(panel (e) of Fig. 2).
F(λ)=[F (λ)−F (λ)]+F (λ)exp[−τ(λ)], (2) If the residual flux is not taken into account, then the one
total cloud cloud
component model – Lorentzian (b) or Gaussian (c) profiles
here τ(λ) istheopticaldepthofthecloudatthewavelength – is not adequate, the reason being that a one-component
λ. The line flux residual (LFR) is the fraction of the QSO Voigt profile cannot describe the unsaturated bottom and
fluxwhichisnotcoveredbythecloud.Thesedefinitionsare far wings of the line simultaneously. Using additional com-
(cid:13)c 2015RAS,MNRAS000,1–13
4 Klimenko et al.
a b c d e
x
u
Fl
d 1
e
aliz 2=2.2 2=89.1 2=29.6 2=2.1 2=2.0
m
AICC = 85.8 AICC = 3299.4 AICC = 1100.9 AICC = 198.3 AICC = 83.0
r
o logN=18.2 logN=18.2 logN=15.1 logN=15.8 logN=18.2 10%
N
0
-30 0 30 -30 0 30 -30 0 30 -30 0 30 -30 0 30

v, km/s
Figure 2. Panel (a) shows a single-component H2 absorption line with logN = 18.2 and b = 4kms−1 and a fit by a one-component
model(redline).Thefluxatthebottomofthelinegoestozero. Weaddaresidualfluxofabout10percent(forclarity)tothedatain
panels(b),(c),(d)and(e).Thefluxatthebottomofthelinesdoesnotgotothezerolevel,andsimultaneouslythelineshaveLorentzian
wings.Asimpleone-component modelcannot describe thisline(cases (b)and(c)) andreturns alarge reduced χ2.However, ifweadd
a new unresolved component to the model, the χ2 will be significantly decreased (case (d)) but not the statistical criterion, the AICC
(seeKingetal.2011),andthereturned columndensitywillbetwoorderofmagnitudetoosmall.

Table 2. Results of the previous analyses of the H2 system at
z=2.811towards Q0528−250.

Year logNtot NComp. Resolution Ref. J=0 J=1
L1R0 L1R1
1985 16.46±0.07 1 [1]
1988 18.0 1 10000 [2] ux
1998 16.77±0.09 1 10000 [3] d fl
22000065 1188..4252±+−000...110372 –2 40000 [[45]] malize J=2 J= 3
2011 16.56±0.02 3 45000 [6] Nor L1R2 L4P3
2015 18.28±0.02 2 45000 This work
[1] Levshakov&Varshalovich(1985);[2] Foltzetal.(1988);

[3] Srianand&Petitjean(1998);[4] Srianandetal.(2005); J=4 J=5
[5] Ćirkovićetal.(2006);[6] Kingetal.(2011) L5P4 L4P5

Velocity, km/s
ponents, as shown in panel (d), a result with a satisfactory Figure3.ExamplesofH2lineprofilescorrespondingtothetran-
χ2 is obtained. However, this solution is incorrect, because sitionsfromJ=0−5levels.Twocomponentsareclearlyseenin
theresultingcolumndensity(logN =15.8) ismuchsmaller the J=4 and 5 lines. We have found that two components are
than the input one (logN = 18.2). To distinguish between enoughtoobtainasatisfyingχ2.Theoriginofthevelocityscale
cases (d) and (e) we propose a new method based on an istakenattheredshiftoftheH2 component, zA=2.81099.
analysis of several absorption lines with different oscillator
strengths. The description of the method and application
to the analysis of H in Q0528−250 are described in more
2
detail later.
4.1 Number of components
The profiles of the H absorption lines have a com-
2
plex structure that cannot be fitted with a single com-
ponent. At least two components are clearly seen in the
lines corresponding to J=4,J=5 rotational levels (see
4 MOLECULAR HYDROGEN
Fig.3). Since the first identification of the H system in
2
Molecular hydrogen lines are detected in the spectrum of the quasar (Levshakov & Varshalovich 1985) other studies
Q0528−250 from the DLA at z =2.811. Column den- have been conducted, providing discordant results (see Ta-
abs
sityofneutralhydrogeninthisDLAsystemislogN(Hi)= ble2).Srianand et al.(2005)used atwo-componentmodel,
21.35±0.07 (Noterdaeme et al. 2008). H lines correspond whileKinget al.(2011)pointedoutthatathree-component
2
to transitions from rotational levels up to J=5. To fit the modelisverystronglypreferredovertwo-componentmodel.
molecular hydrogen lines, the spectrum has been normal- Kinget al. (2011) used a fitting procedure where they in-
ized with a continuum constructed by fitting the selected creased the number of components in the absorption sys-
continuum regions devoid of any absorptions with spline. tem in order to minimize the corrected Akaike information
(cid:13)c 2015RAS,MNRAS000,1–13
Partial covering of Q0528−250 by intervening H clouds. 5
2

1,0
x
Flu 0,5
d
e
z
ali L1P1 L1R0 L2P1 L3P1
m
0,0
or
N -30 0 30 -30 0 30 -30 0 30 -30 0 30
0,05

0,00
0 10 0 10 0 10 0 10

1,0 W1Q4
x
Flu 0,5
d
e
z
ali L4P1 L4R0 L9R1 L9R0
m 0,0
Nor -30 0 30 -30 0 30 -30 0 30 -30 0 30
0,05

0,00
0 10 0 10 0 10 0 10

Velocity km/s
( )
Figure4.SomeoftheabsorptionlinesofH2 fromtheJ=0,1levelsdetectedinthespectrumofQ0528−250andthebestVoigtprofile
fit (red line). Two components of the absorption system are shown by blue dashed lines. Only unblended saturated lines are present.
Thepresence ofprominentLorentzian wingsofthelinesindicatehighH2 columndensity. Theadditional panelsshow aclose-upofthe
bottomofthecorrespondinglines.Thedashedblackhorizontallinerepresents thezerofluxlevel.Itisclearlyseenthatabout2percent
of the LFR is present at the bottom of the saturated H2 lines. The x-axes show the velocity offset from the centre of H2 component
at z=2.81099. In the profile of H2 lineL9R1, some pixels inthe bottom of the lineare lower than 2 per cent level, which is probably
becausethislineisblendedwithLyαforestabsorptionlines.
criterion, AICC2 (Sugiura 1978; King et al. 2011). The re- at the bottom which significantly differs from the zero flux
sulting total column density differs by two order of magni- level (see Fig. 4 and Fig. 5). As a consequence, the fit by
tudefromSrianand et al.(2005).Inthatcaseacriterionfor Srianand et al.(2005)givesthelargereducedχ2.Itisprob-
choosing preferred model is the consistent derived physical able that the signal-to-noise ratio in the previous spectrum
parameters of a cloud. It can be noted note that by using was insufficient to detect theresidual flux.
H columndensityreportedbyKinget al.(2011)(basedon Thecurrentstudyshowsthattheresidualfluxdetected
2
the three-component model) a N(HD)/2N(H ) ratio is ob- in the bottom of saturated H lines in the spectrum of
2 2
tained,thatisaboutanorderofmagnitudehigherthanthe Q0528−250 istheresult ofapartial coverageeffect.Inthis
primordial one. Meanwhile, the H column density in two- case, the profiles of H lines can be very well fitted by two-
2 2
component model of Srianand et al. (2005) gives a reason- component model. Therefore, there is no need to increase
ableN(HD)/2N(H )ratio,whichisconsistentwiththetyp- thenumberofcomponentsinH profiles(suchasit isdone
2 2
icalvaluesmeasuredathighredshift(Balashev et al.2010). byKing et al.2011)toexplainthecomplexstructureoflines
A very important point for the choice of a reasonable (anexampleisgiveninFig.2).Inthenextthreesubsections,
absorption profile model in the case of Q0528−250 is the we provideevidenceof the existenceof thepartial coverage
presence of Lorentzian wings in J=0, J=1 line profiles in the spectrum of Q0528−250.
(see Fig. 4). It is an indicator of the high H column den-
2
sityofabsorptionsystemwithlogN(H )>18whichiscon-
2
sistent with the result reported by Srianand et al. (2005).
4.2 Zero-flux level correction
However, in the new spectrum of Q0528−250 the H lines
2
with prominent Lorentzian wings have some residual flux TomeasuretheLFRinH absorptionlinesweneedtoderive
2
the zero flux level in the spectrum. A non-zero flux in the
coreofasaturatedabsorptionlinecanbetheresultofinac-
2 This statistical criteria allows for a choice of prefered model curatedeterminationinbetweenspectralordersofscattered
among several models with different numbers of fitting parame- light inside the instrument. The zero flux level in the spec-
ters.AICC=χ2+2p+2p(p+1),wherepisthenumberoffitting trum can be estimated using the saturated Lyα absorption
n−p−1
parameters, n is the number of spectral points, included in an lines which are numerous and almost uniformly distributed
analysis. overthewavelengthrangewhereH2 absorptionlinesarelo-
(cid:13)c 2015RAS,MNRAS000,1–13
6 Klimenko et al.
Thecomparison oftheobtainedresidualfluxintheJ=0,1
1,0 H lineswith thelevelof theresidualfluxintheLyαforest
0,10 2
linesisshowninFig.5.Thefilledsquaresrepresenttheresid-
0,8 ual flux in H2 lines versus its location in the spectrum. We
0,05 LFR haveestimatedthelinefluxresidualatthelevel2.22±0.54%
x
zed flu 0,6 0,00 olefvtehl.e continuum, which significantly exceeds the zero-flux
mali -8-6-4-20 2 4 6 8 Non-zero residual flux at thebottom of saturated lines
or 0,4 can also be the result of the convolution of the satu-
N
rated lines with the instrumental function, or an imperfect
Width
data reduction, and/or of the blend of several unresolved
0,2
unsaturated lines. First, however H lines are wide with
2
abs = 3720.5 ¯ widths larger than the FWHM of the UVES spectrograph
0,0 (6km/s,i.e.∼0.08Å), so after convolution with the instru-
ment function, the flux at the bottom of these lines must
-30 -20 -10 0 10 20 30 still go to zero. Secondly, the improper data reduction is
Velocity, km/s not a viable explanation of the residual fluxes of saturated
Figure 6.Anillustrationof the residual flux measurement pro- Lyα lines are consistently equal to zero. Also we tested the
cedure. To estimate the residual flux we determine the median dependence of residual flux on the width of the bottom of
fluxforpixels inalineprofile thatare intherange of±1σ from
line for H and for the Lyα forest lines. This is shown in
2
the flux at the nearest pixel to the central wavelength λc. This Fig.B1. Several lines in the Lyα forest were detected; these
region isshownbypurple shading. Werefer tothe widthofthis
havesimilar widthsastheH linesandgotothezerolevel.
regionasthewidthofthebottomoftheline. 2
The profiles of these lines are shown in Fig.B2.
Lastly,theresidualfluxcannotbearesult ofthecom-
positionofseveralunresolvedunsaturatedlinesbecausethe
cated.Lyαlinesareassociatedwithintergalacticcloudsthat
lines haveLorentzian wings.
are larger than several kpc, thus it is most likely that they
coverthe background source completely.
Wideflatbottomlines(∆λ>1Å,i.e.δv>80kms−1) 4.3.1 The λf test
wereselected,whichguaranteesthatlinesaresaturatedand
Anadditionaltestwhichconfirmsthepresenceofthepartial
thereforemeasuredfluxesinthebottomofthelines(LFRs)
coverage effect is discussed here (and illustrated in Fig.7).
aretherealzerofluxlevelinthespectrum.Theselineswere
Molecular hydrogen lines from J=0,1 levels have very dif-
selected in the spectral region 3500−4700Å. To estimate
ferent values of the product λf (from 0.6 for L0P1 to 36.0
the residual flux at the bottom of the line the procedure
for W1Q1). It is known that the flux in the bottom of an
illustrated in Fig.6 was implemented. We selected several
absorption line decreases exponentially when λf increases,
pixelsinalineprofileforwhichfluxiswithinfc±1σi,where F ∝ exp(−λf). The results of the calculation of this de-
fc is the flux at the centre of the line and σi is the error in pendence for an absorption line in the spectrum with VLT
pixel i. The residual flux was calculated as the median of resolution (FWHM=6kms−1) are shown in Fig.7 by the
thefluxintheselected pixels.Thewidthofthelinebottom
dashedlines.Forsimplicity,themodeledlineprofileconsists
was estimated as the differencebetween theright most and ofonecomponent.TheDopplerparameterb=4kms−1and
left-mostselectedpixelsfromthelinecentre(seeFig.6).The
top-leftpanelofFig.5showstheLFRvaluesobtainedforthe theDampingwidthΓlm isthesameasfortheH2 lineL2P1.
Because the equivalent width of the line gets to the log-
selectedsaturatedLyαabsorptionlines.Theaveragevalueis
arithmic part of the curve of growth, we do not consider
foundtobe(−0.21±0.04)%.Thestandarddeviationofthe
thedifferentDopplerparameters.Twodashedcurves(violet
points is σ ≃ 0.22%. The green stars show the LFR values
and grey) are calculated for the line with column densities
measuredatthebottomoftheLyαandLyβlinesassociated
logN =15and16.Thecurveisshiftedfromrighttoleft as
with the DLA-system at z =2.79 and with the Lyα line
abs the column density N increases. For a higher column den-
of the second DLA-system at z =2.14. These lines are
abs sity, the residual flux equals zero for a wide range of λf
themost saturated lines for thespectrum.
(becausethelineissaturated)anddiffersfromzeroonlyfor
small values of λf 6 5 (the case of an unsaturated line).
In a case of a multicomponent line (i.e. the blend of two or
4.3 Partial coverage of H absorption lines
2 threelineswithsmallcolumndensitylogN =15),theresid-
ToestimatetheresidualfluxinH linesweselectedJ=0,1 ualfluxinthebottomoftheblendedlinewouldalsobehave
2
lines without apparent blends. The LFR was measured by exponentially(dashed-dotcurves).However,thebehaviorof
the same technique as for the Lyα forest lines. The analy- theresidualfluxesinthebottomofthesaturatedH linesin
2
sis found that the flux in the bottom of the lines is quite thespectrum of Q0528−250 isquitedifferent.Theresidual
constant over large velocity range δv ≃ 10kms−1 (i.e. the fluxes in the H lines are shown by the filled blue and red
2
dispersionofpointsiswithintherangeoftheaveragestatis- squares. The squares do not follow the expected behavior,
tical error) that is wider than thefull width half maximum moreoverthepointsscattersimilarlyaroundamedianvalue
(FWHM) of the UVES (6kms−1) and is comparable with in the whole range of λf.
theshiftbetweencentresofthetwocomponents.Therefore, Twomodelswereappliedtoobtainaconsistentandcor-
thestructureofH systemhasnoeffectontheresidualflux. rect fit to the H lines. Model (i) considers the best fit of
2 2
(cid:13)c 2015RAS,MNRAS000,1–13
Partial covering of Q0528−250 by intervening H clouds. 7
2

L14 L12 L9 L7 L4 L2 L1 L0
4 4
% %
R, R,
F2 F2
L L
-0.21 – 0.22 2.22 – 0.54
0 0
Ly CIII Ly ArI FeIII
-2 -2
3600 3800 4000 4200 4400 4600 3600 3800 4000 4200

1 1

0,08 0,08
0,06 0,06 L4P(1) L4P(1) L9R(0)
0,04 0,04
0,02 -0.22% 0,02 2.4%
0,00 0,00 2.4 % 1.7 %
0 -0,02 4254,8 4255,2 4255,6 4256,0 -0,042005,3 4005,4 4005,5 4005,6 4005,7 0
4156 4157 4158 4252 4256 4005 4006 3778
Waveleng th (¯) Wavelen gth (¯)
Figure 5.Topleft:Analysisofthezerofluxlevel inthespectrum ofQ0528−250. TheLFRsmeasured atthebottom ofthesaturated
absorption lines in the Lyα forest are plotted against the wavelength as filled squares. The residual fluxes are expressed in per cent of
thecontinuum.ThegreenstarsshowtheLFRmeasuredintheLyβ andLyαabsorptionlinesoftheDLAsystematzabs=2.79andthe
Lyα absorption from the DLA system at zabs=2.14. The average value of the zero flux level is shown by the red horizontal line. The
scatter intheLFRatthebottom ofthelines(i.e.noiseinthespectrum) isfoundtobeatthelevelof0.22percent. Bottom left:afew
examplesofsaturated Lyαlines.Thethirdbottom panelisthezoomofapartofthesecondpanel.Topright:theresidualfluxesatthe
bottom of the H2 absorption linesfrom rotational levels J=0(blue) and J=1 (red) are plotted against wavelengths as filledsquares.
The average value of the LFR in the H2 lines is shown by the blue line. Blue circles and orange horizontal lines show the positions of
QSObroademissionlinesandtheirwidthstakenfromVandenBerketal.(2001).Bottomright:AfewsaturatedH2absorptionlinesare
present. The red dashed lines show the estimated residual flux in each H2 line. The black line indicates the zero level. The first panel
shows the zoom of a part of the second panel. It can be seen that the flux in the pixels located at the bottom of saturated H2 lines
systematicallydeparts fromzero. Theshiftisabout2percentofthecontinuum.
the certain H line (e.g. L2P1), which has a value of λf = 4.4 Voigt profile Fitting
2
4.23 and LFR=1.7%of thetotal flux.Thelinehasawide
flat bottom (δv ∼ 10kms−1) and Lorentzian wings. To fit A Voigt profile fitting of the H2 absorption lines was per-
formed,takingintoaccountthepartialcoverage.Todescribe
this line without the partial coverage, it is necessary to use
a complex structure of line profiles we have divided the to-
several unsaturated components in the line profile. There-
talfluxdetectedbyanobserverintotwoparts:fromamain
fore, this model describes the line profile well. Model (ii)
source and an additional one. In addition, we consider the
takes into account the partial coverage, and the line L2P1
H system to be composed of two components (A and B)
canbefittedbytheone-componentmodelwithhighcolumn 2
at redshifts z =2.81099 and z =2.81112. The light from
density log N=18.2 and LFR=1.7% (theblue line in the A B
the main source (∼98% of the total flux) is intercepted by
right-handtoppanel).Thismodelalsodescribesthelinepro-
the two H components and does not produce any residual
file accurately. Usingonly oneH line we cannot determine 2
2 flux in the saturated lines. The light from the additional
the most probable model. However, if we consider several
source that passes by H clouds is not absorbed and there-
H lines from the same J level which cover a wide range of 2
2 fore produces a uniform residual flux in H lines. The flux
λf, we will be able to discriminate it. It is known that the 2
in the absorption line of two components A and B can be
lines from thesameJ levelcorrespond tothesamephysical
described as
regionofamolecularcloud;thereforethelinesaredescribed
bythesameset of physicalparameters(thecolumn density F(λ)=(1−fc)Ftotal+fcFtotale−τA(λ)e−τB(λ), (3)
N and Doppler parameter b). The difference between line
profiles originating from the same J level is caused only by wherefc(inrelativeunits)isthecoveringfactorforH2lines.
However, because the physical conditions in the A and
thedifferentvaluesofλf forthelines.Therefore,thecorrect
B clouds (such us linear size, volume density, etc.) might
model of an absorption system must describe all measure-
be different, we can expect the covering factors of quasar
mentsof residual fluxin thebottom of lines from a given J
emission regions by two H clouds also to be different. In
levelsimultaneously.Usingthebestfitparametersformod- 2
this case the construction of the H line profiles is more
els (i) and (ii) we have calculated the dependence of the 2
complicated and we present an analysis of this case below
residualfluxonλf (thickredandbluecurves,respectively).
(see AppendixA). Here, it is important to note that taking
The thick red line cannot describe all squares in the main
intoaccounttwocoveringfactorsdoesnotallowforabetter
panel simultaneously, whereas thebluethick line can.
fit to theH lines (see thediscussion in AppendixA).
2
Then the absorption lines for each J level were de-
(cid:13)c 2015RAS,MNRAS000,1–13
8 Klimenko et al.

m
u
lo lti-c log L2P1
g o
m N
N
po =
= n
% 1 en 15
6 t lin
R, e
1.7%
F
L
v, km/s
fit with additional source of intensity
fit by 3 component model
f, ¯f
Figure7.TheresidualfluxatthebottomoftheH2 absorptionlinesisgivenversustheparameterλf (wheref istheoscillatorstrength
of a line). Blue and red points correspond to the H2 lines from the J=0 and J=1 levels. The theoretical flux at the bottom of the
one-component linewithafixed columndensityN and Dopplerparameter bisshown bythedashed curve. Thevioletand grey curves
correspond tocolumndensities logN(H2)=15and16. Itisseenthat allpoints cannot bedescribed byone curve simultaneously. The
blue dash-dotted linerepresents the calculated residual flux at the bottom of a composite line,which consists of two components with
logN(H2)=15forbothlinesandavelocityseparationofδv=2kms−1.Thegreen dash-dottedlinepresents thesamemodelforthree
components with the same column densities and velocity separations of δv = 2 and 4 kms−1, respectively (see text). An example is
showninthe right-handtop panel. Wefitacertain H2 line(L2P1)bytwomodels:onecomponent withthe LFR(blue line)andthree
components (redline).Bothmodelsgiveadequate fitoftheline.ButmodelwithouttheLFRcannotdescribefluxinalllinestogether,
as shownbythe thick red lineinthe graph. In contrast, all points are described by aone-component model withresidual flux(see the
thickblueline).
scribed using seven fitting parameters: z , z , b , b , N , the current analysis, we selected H lines that are free of
A B A B A 2
NB, fc. We used uniform values of fc over the whole wave- anyobviousblends.Thesampleexaminedcontains99lines.
length range, because the residual flux in the H lines ThebestfitofH linesareshowninFig.C1−Fig.C5ranked
2 2
is almost independent of the wavelength (see Fig.5). The following wavelength positions. The best-fittingparameters
Doppler parameter b is a function of the rotational level areillustrated inTable3.Thereducedχ2 is1.08 (thenum-
J. To estimate fitting parameters Markov Chain Monte berof fittingpoints ∼1500).
Carlo (MCMC) method was implemented, and to speed TheH absorptionsystemishighlysaturated.Thetotal
2
the convergence the Affine invariant ensemble sampler by H column densities are 18.10±0.02 and 17.82±0.02 for
2
Goodman & Weare (2010) was used. The main advantage the A and B components consequently. This is consistent
of this searching algorithm is tobetterexplore a parameter with the presence of the Lorentzian wings in the profiles
space and to avoid using the partial derivatives of the χ2 of J=0 and J=1 levels. The obtained orto–to–para ratios
function which eases a numberof numerical issues. This al- are 2.7±0.1 and 3.2±0.3 for the A and B components,
lowed for more reliable estimates of the fitting parameters respectively.Thecorresponding kinetictemperaturesof the
in comparison with other algorithms. H2 clouds are T01,A =141±6K and T01,B =167±13K.
Fig.8showsacomparisonoftheH excitationdiagrams
2
for ourmeasurementsand those of previouswork.The left-
4.5 Fitting results
hand panel shows the result of the analysis performed by
The H absorption system at z =2.811 has more than 130 Kinget al. (2011), the right-hand panel shows the result
2
absorption lines from J=0 to J=5 rotational levels. For from the present work. Because the results presented here
(cid:13)c 2015RAS,MNRAS000,1–13
Partial covering of Q0528−250 by intervening H clouds. 9
2
Table 3. Column densities and Doppler parameters obtained Table 4. Best fitting parameters for HD molecular lines in the
from Voigt profile fitting of the H2 system at zabs=2.811 to- spectrum ofQ0528−250.
wardsQ0528−250aftertakingcareofpartialcoverage.
zabs logN(cm−2) b(kms−1)
System J zabs logN(cm−2) b(kms−1)
2.811121±0.000002 13.33±0.02 2.25±0.53
A 0 2.8109950(20) 17.50±0.02 2.66±0.05
1 2.8109950(20) 17.93±0.01 2.71±0.05 HD/2H2 (1.48±0.10)×10−5
2 2.8109952(5) 16.87±0.03 2.75±0.03
3 2.8109934(5) 15.97±0.07 2.87±0.07
ing improved laboratory wavelengths, Ivanov et al. 2010).
4 2.8109938(8) 14.18±0.01 4.79±0.11
The presence of HD lines in this system was first re-
5 2.8109938(8) 13.58±0.02 5.04±0.39 ported by King et al. (2011). The HD molecular lines are
present only in thecomponent B. Some of theHD lines are
B 0 2.8111240(20) 17.16±0.03 1.17±0.06 shown in Fig.9. Wehaveestimated the HDcolumn density
1 2.8111230(20) 17.67±0.02 1.14±0.06 logN(HD)B = 13.33 ± 0.02 by analysis of the two most
2 2.8111235(7) 16.64±0.03 1.22±0.02 prominent unblended absorption lines, L4-0R(0) and L8-
3 2.8111238(6) 16.24±0.06 1.25±0.03 0R(0). The results of Voigt profile fitting are presented in
Table4. Other HD lines are highly blended and cannot be
4 2.8111231(6) 14.20±0.01 1.72±0.09
used in the analysis. The obtained column density is sig-
5 2.8111231(6) 13.60±0.02 2.38±0.45
nificantly less than that required to produce self-shielding,
logN(HD) ≃ 15, thus we can set only a lower limit to the
isotopic ratio D/H in the cloud. Because the total column
are close to data reported by Srianand et al. (2005), these
density of H in the component B is logN(H ) = 17.85±
arenotshownhere.WeshowexcitationdiagramsfortwoH 2 2
2 0.02,weestimateD/H>N(HD)/2N(H )=(1.48±0.10)×
componentsatz =2.81099 (2.811001 foundbyKing et al. 2
A 10−5. The obtained HD column density is close to the re-
(2011)) and z =2.81112. The third H component found
B 2 sult, logN(HD) = 13.267±0.072, reported by King et al.
by King et al. (2011) at z =2.8109346(11) is represented
C
(2011).However,takingintoaccountthesignificantlylarger
by the green stars. Although the ratio of J=0 and J=1
H column density of the component B we have obtained a
levels of third component is the same as in other compo- 2
lowervalueofN(HD)/2N(H )thantheresultbyKing et al.
nents,theexcitationdiagram isnotphysicallyrealistic (the 2
(2011). This limit is consistent with D/H ratio obtained
excitation temperatures T and T for the third compo-
02 13 from the analyses of atomic species in quasar spectra (e.g.
nentarenegative).Thismightbetheresultoftheincorrect
Oliveet al. 2012). The comparison of this result with other
model being used. It is seen, that the discrepancy between
N(HD)/2N(H ) measurements at high redshift is shown in
ourresults and thoseof King et al. (2011) is larger only for 2
Fig.10. The N(HD)/2N(H ) in this system is consistent
the low J levels, where theinfluence of the partial coverage 2
withothervaluesandcorrespondtopredictionsofdeuterium
effect on the structure of the line profiles is significant. For
chemistrymodelsofdiffuseISMclouds(e.g.Balashev et al.
high J levels, where column densities of H are less, the re-
2 2010; Liszt 2014).
sults agree in 1σ. To sum up, the total column density of
Note that in the component B we detect HD and Ci
H from the measurement is 18.284±0.025, that is about
2
whereas in the component A these species are not present,
two orders of magnitude larger than the value reported by
despite the higher H column density of this component in
King et al. (2011) 16.556±0.024. 2
comparison with the component B. As for the component
It should be noted, that the same values of N and b
A we set an upper limit for HD and Ci column densities
parameters for all transitions of one J level were used. Us-
which are logN(HD) 6 13.1 and logN(Ci) 612.0. The
ingthemodelwith takingintoaccountthepartial coverage A A
lackofHDandCiinthecomponentAmightbetheresultof
effect we obtain the reduced χ2 ≃1 without an increase of
higherlocalUVradiation(e.g.brightyoungstarsnearcloud
thestatisticalerrorsofthespectrum.Thisisimportant,be-
A).ItcandestroyHDandionizeCiwhiletheH molecules
cause in the previous analysis of H system in Q0528−250 2
2
self-shieldagainstthelocalstellarradiationbecausetotheir
King et al. (2011) noted that without artificially increasing
high column density.
the statistical errors, the reduced χ2 was ≫1 (see the cap-
tion of table6 in King et al. 2011).
The value of the residual flux in H lines is fitted as
2
an independent parameter of an analysis. The best value 6 DISCUSSION
is 2.40±0.07% of the continuum, which agrees with the
The presence of a residual flux at the bottom of the
averagevalueoftheresidualfluxobtainedfromtheanalysis
saturated lines of the H system at z =2.811 towards
of J=0,1 H lines (see Section4.3). 2 abs
2 Q0528−250 can be interpreted according to the arguments
described in thefollowing subsections.
5 HD MOLECULES
6.1 Unresolved multicomponent quasars
Inthenewspectrum(obtainedbyco-addingallpreviousand
additional new observations) the molecular HD lines asso- At redshift z∼2, the internal structures in the emitting
ciated with the H absorption system were detected (us- regions of of quasars with transverse dimensions . 3kpc
2
(cid:13)c 2015RAS,MNRAS000,1–13
10 Klimenko et al.

King et al. (2011) This work
2.8110056(9) 2.8109950(20)
2.8111223(7) 2.8111240(20)
+6
gJ 2.8109346(11) T01=141-6 K
/
NJ +55
g T01=123-55 K
o +45
L T01=111-45 K
+65
T01=134-65 K
+13
T01=167-13 K
-1 -1
E(J), cm E(J), cm
Figure 8.The excitation diagram forH2 towards Q0528−250. Here, NJ is thecolumn density ofthe transition from the J level with
gJ degeneracy. Solidlines(blue,redandgreen)correspondtotheexcitationtemperature T01 derivedfromJ =0andJ =1levels.The
left-hand panel shows the result of the analysis performed by Kingetal. (2011). The excitation diagrams for different components of
theH2 systemareshownbyblue,redandgreencolors.Theright-handpanelshowstheresultsfromourpresentwork.Instead ofthree
unsaturated components weusetwocomponents withhigherH2 columndensities.

1,0 Q 0528-250 (King et al. 2011) log N(HD
0,8 -4 ) = 15
J 2123-0050
J 1232+082
ormalixed flux 01,,60 L440-107R,8(0) 3L957-04R,2(0) 3854,5L8 -0R(0) 383W80,5-0R(0) HD)/2(H)N2 -5 D/H atomicQ 0528-250 J 1237J+ 10463497+1117 J 1331+17J0 A0812+3208A
N (N (This work)
0,8 W1-0R(0) og J2100-064 J 1331+170B
l
L11-0R(0) L14-0R(0) L16-0R(0)
0,6 -6
3762,6 3749,9 3657,6 3602,5
Q 0812+3208B
Wavelength, ¯
Figure 9. The line profiles of HD molecules in the absorption
17 18 19 20 21
systematzabs=2.811inthespectrum ofQ0528−250. HDlines
aredetectedincomponentBonly.ThefittotheW0-0R(0)lineis log N(H2)
inconsistentowingtoapoordefinitionofthelocalcontinuumnear
Figure 10. The measurements of N(HD)/N(H2) vs N(H2) in
thebaseoftheLyαlineofthesecondDLAsystematzabs=2.14. absorption systems at high redshift. The data are taken from
Ivanchiketal. (2015). The estimates of N(HD)/N(H2) toward
Q0528−250 fromthis workandKingetal.(2011)are shownby
redandbluecircles,accordingly.Thebluehorizontalstriprepre-
are unresolved by UVES observations. For example, bi- sents the ratio of atomic Diand Hi measured inquasar spectra
nary quasars with separations of the order of ∼10kpc (Oliveetal.2012).Thesoliddashedlinecorrespondstoconstant
(Hennawiet al.2006;Vivek et al.2009)mightremainunre- columndensityofHD,logN(HD)=15.
solved. Partial coverage can arise ifQ0528−250 has acom-
plex multicomponent structureand if not all of thecompo-
nentsare covered bytheabsorbing H clouds.
2
TheavailableVeryLongBaselineArray(VLBA)image
of Q0528−250 (Kanekaret al. 2009; Srianand et al. 2012) dio emission core component is 65×380pc (Kanekaret al.
showsanunresolvedcomponentcontaining∼94percentof 2009), that is significantly larger than the size of the H
2
thetotalfluxintheradioband(seetable6ofSrianand et al. clouds.WehavealsolookedattheimagesofPKS0528−250
2012).Another∼6percentisprobablyemittedbyadiffuse in 13 and 4cm taken as part of National Radio Astron-
component.However,itisintriguingtonotetheconsistency omy Observatories’ VLBA calibrators. The sources are un-
of these numbers with our findings. The spatial size of ra- resolved even at higherresolution achieved in theseimages.
(cid:13)c 2015RAS,MNRAS000,1–13