Now you see them, now you don'tgreatlakesquasars.uwo.ca/media/slides_folder/Runnoe_Mtg... · 2016....

Now you see them, now you don't: The (dis)appearing central engines of changing-look quasars

Jessie Runnoe Great Lakes Quasar Symposium, May 3, 2016

Collaborators: John Ruan, Chelsea MacLeod, Mike Eracleous, Scott Anderson, Paul Green, Eric Morganson, Sabrina Cales, Stephanie LaMassa

Census of known changing-look quasars

๏Serendipitous discoveries (+2)LaMassa et al. (2015), Runnoe et al. (2016)

๏Systematic search of the SDSS (+2)Ruan et al. (2016)

๏Systematic comparison of Pan-STARRS and SDSS (+8)MacLeod et al. (2016) Nic’s talk

๏But… the definition is slipperysearch criteria are not uniform the distribution of variability amplitudes is unknown

A case study: J1011+5442

MJD = 52652, 57073 Runnoe et al. 2016

3000 4000 5000 6000 7000Rest Wavelength [Å]

0

20

40

60

80

f λ [1

0−17

erg

s−1 c

m−2

Å−1

]

[O II][Ne III] Hδ

Hγ

Hβ

[O III]He I

[N II]Hα

[N II]

[S II]


0

20

40

60

80

f λ [1

0−17

erg

s−1 c

m−2

Å−1

]

Photometric variability

19.0

18.5

18.0

17.5

17.0

V [m

ag]

2002 2004 2006 2008 2010 2012 2014Year

19.0

18.5

18.0

17.5

17.0

V [m

ag]

500 days

๏ Top: Raw light curve from

๏ Bottom: grouped and binned light curve

SDSS LINEAR Catalina

from Runnoe et al. 2016

Nebular diagnostics for the narrow-line region

Ruan et al. (2015)

Possible physical interpretations

๏ Nuclear transient: tidal disruption or SNDuration is longer than a TDE and emission lines are a challenge.Too luminous to be SN by 1-2 orders of magnitude.

๏ Obscuration by transiting, dusty thingTransition time is often shorter than the crossing time.Reddening in lines and continuum is inconsistent.

๏ Abrupt drop in the accretion rateInflow timescale is sometimes short enough.Lines broaden as they weaken.


Spectral reddening test

Ruan et al. 2015

๏ The dim-state spectrum is not a reddened version of the bright state.


๏ Obscuration by transiting, dusty thingTransition time is often shorter than the crossing time.Reddening in lines and continuum are inconsistent.

๏ Abrupt drop in the accretion rateInflow timescale is sometimes short enough.Lines broaden as they weaken.


The inflow timescale

tinfl = 0.25h ↵

0.1

i�1�Edd

0.15

��2 h ⌘

0.1

i2 r

10 rg

�7/2 M8

0.35

�yr

๏ Numerical simulations yield inflow timescales an order of magnitude shorter.

e.g., Krolik et al. (2005)

๏ The optical continuum originates at larger radiiSo it would decay much more slowly but……reverberation mapping shows lags compared to the UV continuum (Michael’s talk yesterday) suggesting that it is powered by illumination from the inner disk.


๏ Obscuration by transiting, dusty thingTransition time is often shorter than the crossing time.Reddening in lines and continuum are inconsistent.

๏ Abrupt drop in the accretion rateLines broaden as they weaken. Inflow timescale is sometimes short enough.


Things we would like to find out

๏Look for the inverse transitions.

๏Learn to select new changing-look quasars effectively.TDSS will be powerful because it will obtain new spectra for thousands of quasars.

๏Get a better census and define the population properties of changing-look quasars.

Distribution transition times and variability amplitudes

a.k.a. Cheshire Cat Quasars (according to D. Schneider)

Broad-line profile changes

6200 6300 6400 6500 6600 6700 6800Rest Wavelength [Å]

0

20

40

60

80

f λ [1

0−17

erg

s−1 c

m−2

Å−1

]


0

20

40

60

80

f λ [1

0−17

erg

s−1 c

m−2

Å−1

]

HαBright State


0

2

4

6

8

10

12

f λ [1

0−17

erg

s−1 c

m−2

Å−1

]


0

2

4

6

8

10

12

f λ [1

0−17

erg

s−1 c

m−2

Å−1

]

HαDim State


0

2

4

6

8

10

12

f λ [1

0−17

erg

s−1 c

m−2

Å−1

]


0

2

4

6

8

10

12

f λ [1

0−17

erg

s−1 c

m−2

Å−1

]

HαDim State


0

10

20

30

f λ [1

0−17

erg

s−1 c

m−2

Å−1

]


0

10

20

30

f λ [1

0−17

erg

s−1 c

m−2

Å−1

]

HαBright State

HβBright state

HβDim state

HαBright state

HαDim state

The power source

−1.0 −0.5 0.0 0.5log([N II] λ6583 / Hα λ6563)

−0.5

0.0

0.5

1.0lo

g([O

III] λ5

007

/ Hβ λ4

861)

Star−forming

SF+AGN

AGN

Runnoe et al. 2016

J1021+4645

MJD = 52614, 56769


0

20

40

60

80

f λ [1

0−17

erg

s−1 c

m−2

Å−1

]


0

20

40

60

80

f λ [1

0−17

erg

s−1 c

m−2

Å−1

]

MacLeod et al. (2016), observed in TDSS

J0159+0033

LaMassa et al. (2015)

5

J0126-0859

Ruan et al. (2016)

J2336+0017

Ruan et al. (2016)

The Time-Domain Spectroscopic Survey

๏ Spectroscopic follow-up for time-variable targets from quasars to M dwarfs.

๏ What is TDSS?

SDSS-IV eBOSS sub-program.

4-6 year optical spectroscopic survey.

260,000 spectra of variable objects.

90% will be initial spectra

๏ 10% repeat observations of interesting objects including 3500+ low-z quasars.

Morganson et al. (2015)

A systematic search of the SDSS

๏ Systematic search for objects which have dramatic changes between quasar and galaxy spectral states.

๏ Selection:

• 2,510,060 objects with CLASS = ‘GALAXY’

• 587,306 objects with CLASS = ‘QSO’

• 117 candidates

• 3 changing-look quasars (2 new)

Ruan et al. (arXiv:1509.03634)J. Ruan

What’s in TDSS?

SDSS J090601.32+485148.7, z = 0.390866

4000 5000 6000 7000 8000 9000 10000Observed Wavelength [Å]

0

20

40

60

80

100

f λ [1

0−17

erg

s−1 c

m−2

Å−1

]

52254559805657556991

What’s in TDSS?

SDSS J111938.02+513315.5, z = 0.107006

4000 5000 6000 7000 8000 9000 10000Observed Wavelength [Å]

0

50

100

150

200

f λ [1

0−17

erg

s−1 c

m−2

Å−1

]

5235357071

The transitions go both ways

1974

1984

Tran et al. (1992)Storchi-Bergmann et al. (1993) Aretxaga et al. (1999) Eracleous & Halpern (2001) Shappee et al. (2014)

Collin-Soufrin (1973)Tohline & Osterbrock (1976)Goodrich et al. (1995)Sanmartim et al. (2014)Denney et al. (2014)Barth et al. (2015)

Wavelength (Å)3200 5600

Osterbrock (1981)

Cohen et al. (1986)

Penston & Perez (1984)

NC

G 4151

MRK 1018

The first changing-look AGN

Collin-Soufrin et al. (1973)

1970

1967

1973A&A....22..343C

NGC 3516

There and back again

1974

1984

1975

1984

Denney et al. (2014)

NGC 7582 MRK 590

Aretxaga et a. (1999)

What causes the changing look?

Dust extinction

Ionizing continuum

Supernovae TDEs

• MRK 1018• MRK 993• NGC 7603• NGC 1087• NGC 7582 (?)

• MRK 590• NGC 2617• MRK 883• NGC 3065

• NGC 7582(?) • NGC 3065 (?)• NGC 7582 (?)

Disappearing broad lines8 MacLeod et al.

Figure 5. Quasars with disappearing BELs. The top panels show J0159+0033, which was discovered in LaMassa et al. (2015). The cyanand purple spectra in the second (right) panel show additional BOSS epochs.

in the lower right subpanels of Figures 4–6, eliminate any un-derlying host galaxy light which is present in both spectraand reveal the SED of the variable component. Aside fromthe features resulting from the BEL change, the |∆fλ| curvesare well-described by a power-law of the form fν ∝ νβ. The

blue curves show the best-fit power-law after masking outthe Hα, Hβ, Hδ, Hγ, Mg ii and HeI BELs and allowing thenormalization to vary. We list the best-fit power-law indicesβ in Table 2, which are based on the spectra listed in the pre-ceding columns. The median for the sample is β = 0.2, with

c⃝ 0000 RAS, MNRAS 000, 000–000

Appearing broad lines Changing-Look Quasars in SDSS 7

Figure 4. Quasars that show evidence for emerging BELs. The light curves in the left panel show the SDSS g-band (black crosses),PS1 g-band (blue triangles), and CRTS photometry (red dots). The right panels show the SDSS and BOSS spectra in red and black,respectively, with epochs indicated by the (color-coded) vertical dotted lines in the light curve. The flux difference |∆fλ| between theBOSS and SDSS spectra is shown in the right lower subpanels, where the blue curve is the best-fit power-law fν ∝ νβ , and the red-dashedcurve is a power-law with β = 1/3.

MJD=52944 to 0.9× 10−17 erg cm−2 s−1 A−1 in 55445. Forthe few reverberation mapped sources with MgII coverage,the Mg ii line is typically unresponsive (e.g., Clavel et al.1991; Krolik et al. 1991; Goad et al. 1993; Cackett et al.2015), meaning that the line flux is constant as the con-tinuum level varies. The result is an intrinsic Baldwin effectwhere the EW decreases with increasing continuum flux in agiven AGN. Since the Mg ii EW remains constant, it seemsthat the Mg ii BEL response in J0225+0030 is more pro-nounced than in typical AGN.

4.4 Continuum Changes

While the majority of highly-variable quasars in our searchdid not exhibit emerging or disappearing BELs, they oftenshowed clear changes in the continuum slope in a bluer-when-brighter fashion, as is expected for quasars9 (e.g.,Schmidt et al. 2012). In the 10 cases presented here, clearcontinuum slope changes were observed along with the BELchanges based on the ratio between spectra.

The difference spectra |∆fλ| = |fBOSS − fSDSS|, shown

9 We do not quantify the color changes for the overall sample of1011 objects, as this is outside the scope of the present work. Forsuch a study, we refer readers to Ruan et al. (2014).

c⃝ 0000 RAS, MNRAS 000, 000–000

There and back againChanging-Look Quasars in SDSS 9

Figure 6. A quasar exhibiting disappearing and reappearing BELs. In the lower left panel, a zoom-in on the Mg ii line shows profile

changes, and the flux deviation Nσ = (f2 − f1)/√

σ22 + σ2

1 between the early BOSS (black, f2) and SDSS (red, f1) spectra is shown in

the bottom panel. The lower right panel is identical to the upper right panel except that the flux difference f3 − f2 between the latestBOSS (purple, f3) and earlier BOSS (black, f2) spectra is shown.

an root-mean-square value 0.4. The red dashed curves showa fν ∝ ν1/3 power law expected for a standard thin disk(Shakura & Sunyaev 1973), and the difference spectra forall 10 objects appear qualitatively consistent with this form.Note that in some cases, |∆fλ| deviates from the power-lawat the shortest wavelengths (e.g., top two panels of Figure 5),but this is likely due to a relatively increased noise level inthe SDSS spectrum.

It is well-known that AGN spectra do not typically fol-low the fν ∝ ν1/3 form expected for a standard thin diskin the optical; only the NIR polarized continuum has beenshown to agree with this prediction (Kishimoto et al. 2003).This is true for our final sample as well (the median β ≈ −1in the high-luminosity state). However, |∆fλ| for all 10 ob-jects follows a power-law that is much more consistent withfν ∝ ν1/3 than the individual spectra. This implies thatthe variable component is related to the thermal AD, whichmight be expected if there was a change in the temperaturestructure of the AD due to a change in viscous heating orirradiating flux.

4.5 Comparison to Previously ReportedChanging-Look Quasars

In Table 3, we list four previously reported changing-look quasars (at z > 0.1), including J0159+0033.J012648.08−083948.0 and J233602.98+001728.7 are not inDR7Q and therefore are not in our parent sample. Thesetwo objects were found by Ruan et al. (2015) in an archivalsearch for objects whose pipeline spectroscopic classifica-

Name z Max(|∆g|) DR7 Found(SDSS J) (mag) QSO? here?012648.08−083948.0a 0.198 0.78± 0.02 × ×015957.64+003310.4b 0.312 1.16± 0.06 ! !101152.98+544206.4c 0.246 1.51± 0.02 ! ×233602.98+001728.7a 0.243 1.05± 0.05 × ×

Table 3. Previously reported changing-look quasars (z > 0.1).aFound by Ruan et al. (2015). bFound by LaMassa et al. (2015).cFound by Runnoe et al. (2015).

tions changed between epochs of spectroscopy. Therefore,their search was sensitive to only the most dramatic changesin spectral state, whereas our variability-based selectionis sensitive to dramatic transitions as well as more subtlechanges in AGN type. However, while J233602.98+001728.7shows enough variability to be included in our sample,J012648.08−083948.0 only shows a 0.78 mag dimming go-ing from SDSS to PS1. This suggests that our variabilitycriterion of |∆g| > 1.0 mag is too restrictive.

The Time Domain Spectroscopic Survey of SDSS-IV(TDSS; Morganson et al. 2015) is obtaining repeat spectrafor quasars showing |∆g| > 0.7 mag variability, and is likelyto find a more complete sample of changing-look quasars.J101152.98+544206.4, discovered by TDSS, is in DR7Q butdoes not have a spectrum in DR12. Had it been observedin SDSS-III, we would have recovered this object since itis 1.5 mag dimmer in PS1 (g = 19.879 ± 0.014 at MJD =55838.170) with respect to SDSS.

c⃝ 0000 RAS, MNRAS 000, 000–000

A few things worth noting

• The search criteria are not mature yet.

• Not defined by Δm, but a change in “look”.

• They may be more prevalent in highly variable systems.

• A good light curve goes a long way.

• The selection effects are not well understood.

• We see all the transitions.

12 new changing-look quasars in 2015

Changing-Look Quasars in SDSS 5

Figure 1. Distribution of the maximum magnitude difference|∆g|max versus time lag |∆t|. The black contours show the dis-tribution for the superset of quasars (DR7Q), which is based onSDSS and PS1 photometry. The white contours are the same butfor the subsample of quasars in S82, which are shown as grey datapoints (one point per S82 quasar). The contours show regions con-taining 5, 10, 25, 50, and 75% of the data. The subsample of DR7Qwith repeat spectra are shown by open red circles, but now show-ing the |∆g|max spanned by the spectroscopic epochs versus thecorresponding time span. For objects with multiple spectra, wechoose the two epochs spanning the largest |∆g| for display here.While the data are clumped into “seasons” in the top panel, thedistribution is smoothed out when switching to rest-frame timelag in the bottom panel. Our final selection is limited to quasarswith repeat spectra with |∆g|max > 1 (indicated by the horizon-tal dashed line). The final sample of 10 objects are plotted as bluecrosses.

Our final sample of changing-look quasars is listed inTable 2, and the redshift distribution is compared to thefull quasar sample in Figure 2. We note that all of our ob-jects are at z < 0.63, but this is potentially a selection effect,as we discuss in Section 5. However, this sample extends therange of known changing-look AGN to z = 0.63 at quasar lu-minosities. In Figure 3, the 5007A [O iii] luminosity is shownas a function of redshift for the full DR7Q sample and our

Figure 2. Redshift distribution for the sample of changing-lookquasars (red histogram), the full sample of 1032 highly variablequasars (dotted histogram), those with repeat spectra (dashedpurple histogram), and the entire DR7 quasar catalog (blue his-togram).

Figure 3. [O iii] luminosity versus redshift for the DR7Q sample(contours), our final sample of 10 (blue crosses), and examplesof previously known changing-look AGN (open circles). For thelatter, the luminosity corresponds to the bright state, cf. Figure 1in LaMassa et al. (2015), whereas the remaining points are basedon the earlier DR7 spectrum using values from Shen et al. (2011).

final objects. Examples of previously studied changing-lookAGN are also shown for comparison.

In the following sections, we compute the flux devia-tion between two spectra at any given wavelength, Nσ(λ) =(f2−f1)/

√

σ22 + σ2

1 (e.g., Filiz Ak et al. 2012), to determinethe significance (in units of σ per spectral pixel) of a BELchange. In particular, we assess the significance of a BELchange by comparing its flux deviation to that of the under-lying continuum at that wavelength. The significance of BELchanges will be higher than that quoted here when Nσ(λ) isintegrated over the pixels spanning the BEL.

c⃝ 0000 RAS, MNRAS 000, 000–000

LaMassa et al. (2015) Runnoe et al. (2016) Ruan et al. (2016) MacLeod et al. (2016)

Black Hole Masses

0.3 - 140108 M⊙

Lbol

*0.9 - 30.2

1044 erg s-1

0.3 - 1.5

Eddington Fractions

*0.1 - 53

10-2 L/Ledd

0.2 - 4

Max. Timescales

4.2 - 9.7years

Physical properties

Now you see them, now you don'tgreatlakesquasars.uwo.ca/media/slides_folder/Runnoe_Mtg... · 2016....

Documents

Transcript of Now you see them, now you don'tgreatlakesquasars.uwo.ca/media/slides_folder/Runnoe_Mtg... · 2016....