Broad Line Variability on Long Timescales

A Complex BLR Response

David Homan

University of the Virgin Islands, 12–07–2017

TH

E

UN I V E

RS

IT

Y

OF

ED

I N BU

R

GH

• Collaborators:Andy Lawrence, Chelsea MacLeod, Nic Ross, AlastairBruce, Martin Ward, Hermine Landt

• BLR response to continuum changesI MgII 2798A: variability rangeI Mrk 110: a tracker for the EUV continuum

1

Different types of MgII behaviour

J224829:

2700 2800 2900 3000 3100 3200Wavelength in QSO Restframe (Å)

5

10

15

20

25

30

35

40

f (

1017

erg

cm

2 s

1 Å

1 ) MJD5264157430

52000 53000 54000 55000 56000 57000Observation Date (MJD)

18.50

18.75

19.00

19.25

19.50

19.75

20.00

20.25

20.50

g-ba

nd M

agni

tude

SDSSPanSTARRS-I

J022556:

2600 2700 2800 2900 3000 3100 3200Wavelength in QSO Restframe (Å)

2

4

6

8

10

12

f λ (

10−

17 e

rg c

m−

2 s−

1 Å

−1) MJD

5294455241554765594556544

52000 53000 54000 55000 56000 57000

Observation Date (MJD)

19.5

20.0

20.5

21.0

21.5

22.0

g-b

an

d M

ag

nit

ud

e

SDSSPanSTARRS-I

MacLeod et al. MNRAS 547 (2016)

2

• New observations for a sample of highly variable AGN:|∆g| > 1, searching for new CLQs• Subsample of 15 objects for which fMgII could be measured

0 1 2 3 4 5 6| f3000|/f3000

0

50

100

150

200|

EW

MgI

I|J002311J004339J022556J022652J081916J100220J111348J132815

J145519J162415J214613J223953J224829J225240J233317

3

• Object J022556 was covered by Stripe 82; total of 15spectra

• Maximum MgII flux change by a factor ∼11

2600 2700 2800 2900 3000 3100 3200Wavelength in QSO Restframe (Å)

2

4

6

8

10

12

f λ (

10−

17 e

rg c

m−

2 s−

1 Å

−1) MJD

5294455241554765594556544

52000 53000 54000 55000 56000 57000

Observation Date (MJD)

19.5

20.0

20.5

21.0

21.5

22.0

g-b

an

d M

ag

nit

ud

e

SDSSPanSTARRS-I

0.0 0.2 0.4 0.6 0.8 1.0

Continuum Flux0.0

0.2

0.4

0.6

0.8

1.0

MgI

I Lin

e Fl

ux

52100

53100

54100

55100

56100

57100

MJD

4

• Line Flux vs. Continuum flux for the entire sample

0.0 0.2 0.4 0.6 0.8 1.0Continuum Flux

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

MgI

I Lin

e Fl

ux

J002311J004339J022556J022652J081916

J100220J111348J132815J145519J162415

J214613J223953J224829J225240J233317

5

Line responsivity in Mrk 110

• Mrk 110: Seyfert I at z=0.0355• Observations over several decades, ∼100 spectra

4000 4500 5000 5500 6000 6500 7000 7500

Wavelength in QSO Restframe (Å)0

500

1000

1500

2000

f (

1017

erg

cm

2 s

1 Å

1 )

52252-SDSS52581-Till53741-Till.54122-Till.57539-WHT57801-WHT

Landt et al. ApJS 174 (2008) 6

• Line fluxes against the continuum at 5100A• Fluxes normalised to a single date

0.0 0.5 1.0 1.5 2.0F5100 (normalised)

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5N

orm

alis

ed F

lux

HHHeII4686HeI5876

Bisschoff & Kollatschny A&A 345 (1999), Kollatschny et al. A&A 379 (2001)7

• Compare normalised line fluxes to HeII 4686 flux• Approximate behaviour with a simple fitting function

0 1 2 3 4HeII 4686 (Ionising Flux)

0.0

0.5

1.0

1.5

2.0

2.5N

orm

alis

ed F

lux

HHHeI 5876

Lawrence et al. (2017, in prep.) 8

• On shorter timescales, the resonsiveness can deviate

• Results from separate RM studies:

0.00 0.25 0.50 0.75 1.00 1.25 1.50He II 4686 (Normalised)

0.0

0.2

0.4

0.6

0.8

1.0

1.2H

Flu

x (N

orm

alis

ed) 51490-51680

52580-5300053000-53140Other

9

• Empirical fitting function

• Apparent cut-off in line responsivity

• Correlation with known lags from RM

0 10 20 30 40RM Lag (days)

0.00

0.25

0.50

0.75

1.00

1.25

1.50Sa

tura

tion

Lev

el

10

Summary

• Range in broad MgII responseI Clear difference with short timescales (RM)

• Large number of spectra allows better analysis for Mrk 110

• The responsiveness of broad lines appears to diminish forhigher EUV continuum values

• Possible connection:I high state versus low state responsiveness

11

12

MgII Line and Continuum flux light curves

52000 53000 54000 55000 56000 57000

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

f MgI

I (er

g cm

2 s1 )

SDSS BOSS

J002311J022556J022652J081916J100220J111348J132815J145519J162415J214613J223953J224829J225240J233317

52000 53000 54000 55000 56000 57000Observation date (MJD)

0.0

0.2

0.4

0.6

0.8

1.0

f c (e

rg c

m2 s

1 Å1 )

13

14 SHEN ET AL.

−50 0 500

5

10

15

% o

f Sam

ple

s RMID 101 RMID 191 RMID 229

0

5

10

15

% o

f Sam

ple

s RMID 267 RMID 272 RMID 320

0

5

10

15

% o

f Sam

ple

s RMID 457 RMID 589 RMID 645

0

5

10

15

% o

f Sam

ple

s RMID 694 RMID 767 RMID 769

−50 0 50

τcent (days)

0

5

10

15

% o

f Sam

ple

s RMID 775

−50 0 50

τcent (days)

RMID 789

−50 0 50

τcent (days)

RMID 840

Figure 8. The CCF centroid distributions (CCCDs) from FR/RSS for the 15lags. The lags are in the observed frame to match Figs. 4–7. The verticaldashed and dotted lines indicate the reported lag and its uncertainties. In allcases there is a reasonably well-defined main peak in the CCCDto determinethe best lag. In a few cases there are sub-structures (possible aliases dueto the sparse sampling of the LCs) in the CCCD that will lead toelevateduncertainties in the lag.

our selection of these detections. In addition, the severalMg IIlags seem to follow the sameR −L relation based on Hβ. Thelatter observation suggests that, at least in some quasars,thereis overlap between the regions in which broad MgII and broadHβ originate (see also Sun et al. 2015), which is expected asboth lines have similar ionization potentials. Of course, moredata and analysis are needed to test this scenario.

The apparent flatness in theR−L relation for our detectionsin Fig. 9 is mostly due to selection effects, and secondly dueto the facts that the statistics and the dynamic range in lumi-nosity for our sample are limited and that the lag measurementerrors are substantial, rendering a potential correlationstatis-tically insignificant. Given the sampling and duration of ourspectroscopic monitoring, we are most sensitive to lags on theorder of tens of days (Shen et al. 2015a). Shorter lags are dif-ficult to detect given the sparse sampling of the spectroscopicLCs, while longer lags are difficult to detect given the limitedtemporal baseline. A more detailed analysis of theR −L rela-tion based on SDSS-RM lag detections is beyond the scope ofthis work, which will require more lag detections and propertreatment of selection biases induced by our program, and will

1042 1043 1044 1045 1046

λLλ(5100 Å) [erg s−1]

1

10

100

BLR

Siz

e [li

ght d

ays]

Bentz et al. (2013, Hβ) Hβ MgII

Figure 9. The BLR size-luminosity relation. Our lag detections are shownas black circles (for Hβ detections) and red squares (for MgII detections).The data for previousz < 0.3 RM AGN compiled in Bentz et al. (2013) areindicated in gray points. Our new lags are consistent with the locations ofthe previous RM AGN used to calibrate the localR −L relation, but are notyet able to constrain theR −L relation independently given the limited num-bers, precision, dynamic range, and possible selection biases inherent to ourprogram (see discussion in the text).

42

43

44

45

46

logL

5100

,AG

N [e

rg s−

1 ]

0.0 0.2 0.4 0.6 0.8Redshift

0 1 2 3 4 5 6 6.8Lookback Time [Gyr]

this worklocal RM AGN

Figure 10. Distribution of objects with detected lags in the redshift-luminosity plane. The red open circles are the 44 local RM AGNcompiledin Feng et al. (2014), and the blue filled circles represent the 15 preliminarylag measurements in this work. Our lag detections probe a newregime in thisparameter space, providing direct SMBH masses over∼ half of cosmic time.

be the focus of future SDSS-RM publications. For example,short lags may be recovered when incorporating more denselysampled photometric light curves, or at least upper limits canbe placed and used in quantifying theR −L relation. On theother hand, simulations can be used to quantify the complete-ness in lag detections as a function of lag, given the parame-ters of the SDSS-RM program (Shen et al. 2015a).

Finally, Fig. 10 demonstrates the improvement of our re-sults in the redshift-luminosity coverage of RM experiments.Our lag detections probe a new regime in this parameterspace, providing RM measurements over∼ half of cosmictime. The median redshift is 0.03 for the local RM sampleand 0.46 for our sample.

3.2. Additional Notes on Individual Objects

While our analysis demonstrates that the majority of theselag measurements are true detections, prior reverberationstudies (e.g., Peterson et al. 2004) indicate that when work-

Shen et al. APJ 818 (2016)14

4000 5000 6000 7000 8000 9000 10000

Observed Wavelength (Å)

0.0

0.5

1.0

1.5

2.0

2.5

f λ (

erg

cm

−2 s−

1 Å

−1)

1e 16

SDSS: 2001WHT: 2016z: ~0.7

J111348

15

χ2-Contours for fit to Hβ

0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4Saturation Level

0.40

0.45

0.50

0.55

0.60

0.65

0.70Sl

ope

Hb

2.300

6.170

11.800

16

Fit results: slope versus lag

0 10 20 30 40RM Lag (days)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Slop

e

17

0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2Normalised Flux

H

H

HeI 5876

Line fluxes for HeII flux above 1.2

18