Giant blob eso

Central Powering of the Largest Lyman-alpha Nebula is Re-
vealed by Polarized Radiation
Matthew Hayes1,2,3 , Claudia Scarlata4,5 , and Brian Siana6

1
Universit´ de Toulouse, UPS-OMP; IRAP; Toulouse, France
e
2
CNRS; IRAP; 14, avenue Edouard Belin, F-31400 Toulouse, France
3
Observatory of Geneva, University of Geneva, 51 ch. des Maillettes, 1290 Versoix, Switzerland
4
Minnesota Institute for Astrophysics, School of Physics and Astronomy, University of Minnesota,
Minneapolis, MN 55455, USA
5
Spitzer Science Center, California Institute of Technology, 220-6, Pasadena, CA 91125, USA
6
Department of Astronomy, California Institute of Technology, MS 249-17, Pasadena, CA 91125,
USA

High-redshift Lyman-alpha blobs1, 2 are extended, luminous, but rare structures that appear
to be associated with the highest peaks in the matter density of the Universe3–6 . Their en-
ergy output and morphology are similar to powerful radio galaxies7 , but the source of the
luminosity is unclear. Some blobs are associated with ultraviolet or infrared bright galaxies,
suggesting an extreme starburst event or accretion onto a central black hole8–10 . Another
possibility is gas that is shock excited by supernovae11, 12 . However some blobs are not associ-
ated with galaxies13, 14 , and may instead be heated by gas falling into a dark matter halo15–19 .
The polarization of the Lyα emission can in principle distinguish between these options20–22 ,
but a previous attempt to detect this signature returned a null detection23 . Here we report
on the detection of polarized Lyα from the blob LAB12 . Although the central region shows
no measurable polarization, the polarized fraction (P) increases to ≈ 20 per cent at a radius
of 45 kpc, forming an almost complete polarized ring. The detection of polarized radiation
is inconsistent with the in situ production of Lyα photons, and we conclude that they must
have been produced in the galaxies hosted within the nebula, and re-scattered by neutral
hydrogen.

The Lyα emission line of neutral hydrogen is a frequently used observational tracer of evolv-
ing galaxies in the high redshift Universe. Lyα imaging surveys typically ﬁnd a large number
of faint unresolved objects and a small fraction of extremely luminous and spatially extended sys-
tems that are usually referred to independently as Lyman alpha blobs (LABs). The compact sources
usually appear to be more ordinary star forming galaxies whereas, since their discovery, much con-
troversy has surrounded the true nature of LABs. Because one of the possible modes of powering
LABs is the accretion of gas from the intergalactic medium15–19 , an eventual consensus about this
powering mechanism holds particular relevance for how the most massive and rapidly evolving
high redshift dark matter haloes acquire the gas required to fuel star formation in the galaxies they
host. Predictions for polarized Lyα radiation were made over a decade ago20 , and have been further
developed to the extent that the polarization signal may act as a diagnostic to study the origin of
the Lyα photons21 . By relying upon the scattering of Lyα photons at large distances from their

1

production sites, these polarization measurements may also provide an independent constraint on
the distribution of neutral gas in the circumgalactic regions, which has proven a diﬃcult measure-
ment to make by traditional techniques. The one previous attempt to detect the polarization of a
LAB returned a null result23 , although very deep observations are required that need considerable
time investments on 8–10 meter class telescopes. In the ESO Very Large Telescope observations
presented here, we observe one of the largest and most luminous LABs on the sky in imaging
polarimetry mode (Figs 1 and 2).

ext
6
21
0.

stack
LAB8
5
21
0.
4
21
0.
3
21
0.
Declination [ degrees ]

2
21

b
0.
1

LAB1
21

ord
0.
0
21
0.

5”
9
20

AGN
0.
8
20
0.
7

a
20
0.

c
6
20
0.

2

1

0

9

8

7

6

5

4
36

36

36

35

35

35

35

35

35
4.

4.

4.

4.

4.

4.

4.

4.

4.
33

33

33

33

33

33

33

33

33

Right ascension [ degrees ]

Figure 1 | LAB1 in Lyα. The large panel to the left shows the combined intensity frame that
has been adaptively smoothed in order to show detail. The Lyα emission is spatially extended
over scales of more than 10 arcsec (80 kpc). The 5 arcsecond scale bar shown in the left panel
corresponds to 40 physical kilo-parsec. North is up and East is to the left. The primary target
of this observation is LAB1 2 , although the position angle was chosen so as to also include
LAB8 4 , which resides some 20 arcsec to the North. The bright point source to the South East
(lower right) is the AGN C112 . Major centres are labelled in the image. The two smaller frames
to the right show the ordinary and extraordinary beams (lower and upper, respectively), which
have been tessellated for quantitative analysis (Supplementary Information). All images are
shown on the same intensity scaling, and even by eye differences in the intensity structure
between the two beams become visible. The cross marks the point that we determine to be
the centre in the subsequent analysis, and is determined as the point of peak Lyα surface
brightness that is not associated with the AGN.

These new, deep polarimetric observations show a substantial fraction of the Lyα emission

2

from LAB1 to be polarized. The data enable the fraction of light that is linearly polarized (P) to
be measured to better than 5% (r.m.s.) in the regions of peak surface brightness, however in these
central regions we find essentially no polarization signal. In order to detect the polarized emission
we need to examine the fainter signal emanating a few arcsec from the centre – when Lyα is emitted
with a high polarization fraction (P ≈ 20%), it is not spatially coincident with the brighter regions
of the nebula. Indeed the highest and most statistically certain measures of P are found some
4–8 arcsec from the centre, forming an almost complete ring of polarized emission. Just a single
resolution element shows a signal-to-noise ratio on P that exceeds 3, but our confidence in these
results is substantially enhanced by the strong degree of spatial correlation between the resolution
elements that are observed to be polarized at high significance. Away from the regions of bright
Lyα emission our data show no measurable polarization signal, but a second nearby extended Lyα
source (LAB84 ) that lies to the north also shows strongly polarized Lyα.

Lyα is a resonant line and photons scatter in neutral hydrogen (H i), undergoing a com-
plicated radiation transfer. At each scattering event, of which there may be millions, a photon
acquires a new direction and frequency, dependent upon the kinematics of the scattering medium
and quantum mechanical probabilities24 . The outgoing frequency determines the optical depth that
the photon then sees25, 26 : small shifts from line-centre (core scatterings) result in effectively no
transfer, while long frequency excursions (wing scatterings) result in significant flights and likely
escape from the system. The more frequent core scatterings may emit polarized radiation but it is
typically with a low fraction (P 7%); the comparatively rare wing scatterings, on the other hand,
may introduce a high degree of linear polarization to the outgoing radiation (P up to 40%)21, 24 . The
observed polarization signal therefore provides an unambiguous signature of the scattering nature
of the diffuse Lyα halos, and distribution of circumgalactic gas. On the atomic level, the detected
polarization signal provides evidence that the Lyα photons are not emerging after a large number
of resonant core scattering events, but are exiting after long shifts in frequency.

The detection of polarization demonstrates that the Lyα photons cannot have been produced
in situ at large radial distances. In order to impart a significant level of polarization on the outgoing
radiation, a geometry is required in which Lyα photons preferentially escape after scattering off hy-
drogen atoms at angles of 90 degrees to their vector prior to the scattering event21, 24 . This geometry
is not expected in the case of wind/shock excitation since thermalized cooling gas will generate
Lyα photons with no preferential orientation or impact angle with respect to the scattering medium.
Similarly, the filamentary streams of cold infalling gas predicted by simulation27, 28 appear to have
too small volume filling factors, and be too narrow (few kpc) for any imparted polarization signal
to be observable17 – a spatial resolution would be required that is currently unattainable by optical
polarimeters. While neither of these rejected scenarios have yet been thoroughly tested by radia-
tive transport studies, the production of polarized Lyα is well understood, and it is difficult to see a
mechanism by which observably polarized Lyα may be produced in these situations. We therefore
conclude that the Lyα photons must have been produced centrally in the galaxies themselves and
scattered at large radii by neutral gas. Even if the halo is accreting intergalactic gas, this process is
not responsible for a large fraction of the Lyα emission.

3

Stokes Q 2σ error P @ conf > 2σ

0.2
0.3

5
0.3

21
0.
0.1

3
21
0.
0.2 0.2
0.0

1
21
0.
− 0.1 0.1

9
5” 5” 0.1 5”
20
0.
20
7 a − 0.2 c e
0.

0.0
Stokes U P with 2σ < 16% χ @ conf > 2σ

0.2
0.3
5
21
0.

P =1
0.1
3
21
0.

0.2
0.0
1
21
0.

− 0.1 0.1
9

5” 5” 5”
20
0.

b − 0.2 d f
7
20
0.

0.0
2

0

8

6

4

2

0

8

6

4

2

0

8

6

4
36

36

35

35

35

36

36

35

35

35

36

36

35

35

35
4.

4.

4.

4.

4.

4.

4.

4.

4.

4.

4.

4.

4.

4.

4.
33

33

33

33

33

33

33

33

33

33

33

33

33

33

33
Right ascension [ degrees ] Right ascension [ degrees ] Right ascension [ degrees ]

Figure 2 | The polarization of LAB1. Scaling is shown to the right of each panel. Panels
(a) and (b) show the Q and U Stokes parameters. Panel (c) shows a map of the polarization
fraction P that we can measure at 95.4% confidence; this error map demonstrates that in
individual resolution elements we can constrain P in the central regions down to 10% (2σ). At
a radius of ≈ 6 − 7 arcsec, this decreases to about the 20% level, and beyond ≈ 10 arcsec
P becomes noise-dominated. Panel (d) shows the value of P in resolution elements where
it can be well constrained; we make a cut based on the error map, and show bins where P
can be measured to better than 16% (2σ). P of around 5–10% is found in the centre, but
increases to 15–20% at ≈ 5 arcsec from the centre towards the South, West, and North edges
of LAB1. P ≈ 20% is also seen around LAB8. Panel (e) shows P in bins for which we can be
at least 95.4% confident of its measurement. The structure around LAB1 and LAB8 remains,
extending to larger radii, and forming a broken ring at P ∼ 20%. Panel (f) shows a smoothed
intensity image with the direction of the polarization vector, χ, shown by red bars. As with panel
(e) only bins showing measurements at more than 95.4% confidence level are shown. There is
a clear trend that, when bins show strong polarization, they orientate themselves tangentially
to the overall structure and local Lyα surface brightness. These features conform well with the
predictions for the polarization of Lyα photons that are centrally produced and scatter at large
radii, or are produced by cooling gas that flows onto the dark matter halo in quasi spherical
symmetry 21 .

The geometrical origin of the polarization signal in the photon scattering model, which is
analogous to Rayleigh scattering, implies also that the angle of the polarization vector χ should
align tangentially to the overall geometry of the system at large radii21, 22 . We also extract this χ
vector from the data (Fig 2, panel f). If we define the geometric centre to be the region of peak Lyα
surface brightness (excluding the active nucleus; Fig 1), these vectors display a strong tendency to

4

orientate along directions tangential to the overall geometry of the system. Again this is in good
agreement with theoretical expectations, should the Lyα photons be produced centrally by ordinary
star-formation and/or AGN-fueled processes and scatter at large radii21 . However unlike the ide-
alized models used for radiative transport simulations, which assume perfect spherical symmetry
and no velocity shear or clumping of the gas, the true Lyα surface brightness is far from smooth
and symmetric. This may reflect either multiple Lyα sources within the halo, large scale inhomo-
geneities in the scattering medium, or both. Polarization of Lyα emission in a clumpy medium has
not yet been tested by simulation although clumping could potentially either suppress or enhance
the polarization21 , and likely depends on the relative size and volume filling factor of the clumps.
Regardless, we would still expect the polarization signature to remain provided that the individual
regions are well resolved, in which case χ would then be expected to align tangentially to contours
in the local surface brightness. This effect is also reflected in the data, and we find polarization
vectors preferentially line up at angles approximately perpendicular to the local surface brightness
gradient at their position. The Supplementary Information includes a quantitative analysis of the
polarization angles compared with both local surface brightness gradients and the overall system
geometry.

There is one final observable prediction that has been suggested by radiative transport simu-
lations. This is the radial profile of the polarization fraction, P, which according to the scattering
shell model should present with P ≈ 0 in the geometric centre and increase with increasing radius.
We have computed this radial profile of P using the same central coordinates as defined previously
(Fig 3) and find strong evidence for P increasing with radius, again in agreement with predictions.
The central point on the profile and most significantly polarized point in the ring are discrepant
at over 2σ. We have fitted the innermost 7 arcsec (5 data points; where data are deemed reliable)
with the simplest function for P varying linearly with radius. Computing the classical p−value,
1 − Γ(ν/2, χ2 /2) where ν is the number of degrees of freedom, shows above 96% confidence in this
hypothesis. The same test performed on a flat profile (P = 8.3%), suggests that a structure in which
P is independent of radius is consistent with the data at 16%. Regardless, the slim possibility of P
that is flat with radius does not present a difficulty for the scattering scenario, which is supported
not only by the detection of polarized Lyα itself, but also by the angular distribution of polarization
vectors; this radial profile is simply the least robust of our constraints. The actual gradient of P
with radius depends upon both the column density and radius of the shell, neither of which we
know – assuming that the idealized geometry holds and that the shell has a radius of 10 arcsec, the
models with column densities of 1019 − 1020 cm−2 would be in agreement with the data (Fig 3). In
this step of the analysis we also compute the total luminosity-weighted polarization fraction within
each radius: within 7 arcsec we obtain an integrated value of P = 11.9 ± 2%. We take this value as
the overall polarization fraction of the nebula, which we quote at 6σ significance.

1. Francis, P. J. et al. A Group of Galaxies at Redshift 2.38. Astrophys J. 457, 490–499 (1996).
arXiv:astro-ph/9511040.

2. Steidel, C. C. et al. Lyα Imaging of a Proto-Cluster Region at <z>=3.09. Astrophys J. 532,

5

Physical distance [ kpc ]
0 20 40 60 80

Intensity, I
0.4
Fraction, P
Radial binning pattern

6
0.3

21
0.
Signal

4
21
0.2

0.
2
21
0.1

0.
0
21
0.
0.0 5”

8
5

20
0.
4

6
20
3

0.

2

0

8

6

4
S/ N

36

36

35

35

35
2

4.

4.

4.

4.

4.
33

33

33

33

33
1 Right ascension [ degrees ]

0
0 2 4 6 8 10 12
Angular distance [ arcsec ]

Figure 3 | The radial polarization profile. Blue points show the fractional polarization, P,
which is measured in concentric 1.5 arcsec annuli. We implement a new binning method
(Supplementary Information) that enables annular measurements of P using independent bins.
The binning structure is shown to the right, the centre the same as illustrated in Fig 1. P has
been corrected for spuriously high polarization fractions that may be measured at low signal-
to-noise29 . In magenta the total intensity profile is shown. Errorbars on both measurements
are 1σ, and the lower profile shows the signal-to-noise ratio of P. In the central regions where
intensity is highest, P is measured at ∼ 4%, although at 1σ is consistent with no polarization.
P and its signal-to-noise both increase with radius; by 5 arcsec P = 10% and is inconsistent
with zero at the 4σ level. P peaks at 18% at 7 arcsec, beyond which radius the noise becomes
dominant. Overlaid are the results of numerical models for scattered Lyα radiation21 in an
expanding H i shell with two column densities: 1019 cm−2 (lower) and 1020 cm−2 (upper). The
curves have been smoothed over lengths of 1.5 arcsec to match our annuli. The actual radius
of the shell is unknown, so the theoretical points have been scaled to shell radius of 10 arcsec,
in order to approximately match both the surface brightness and polarization profiles. This
shows that the predictions are qualitatively matched by observation. Models for scattering
in a static IGM and spherically symmetric accretion are not show since they are not leading
explanations for the powering of LABs.

6

170–182 (2000). arXiv:astro-ph/9910144.

3. Palunas, P., Teplitz, H. I., Francis, P. J., Williger, G. M. & Woodgate, B. E. The Distribution
of Lyα-Emitting Galaxies at z=2.38. Astrophys J. 602, 545–554 (2004). arXiv:astro-ph/
0311279.

4. Matsuda, Y. et al. A Subaru Search for Lyα Blobs in and around the Protocluster Region At
Redshift z = 3.1. Astron. J. 128, 569–584 (2004). arXiv:astro-ph/0405221.

5. Yang, Y., Zabludoﬀ, A., Tremonti, C., Eisenstein, D. & Dav´ , R. Extended Lyα Nebulae
e
at z sime 2.3: An Extremely Rare and Strongly Clustered Population? Astrophys J. 693,
1579–1587 (2009). 0811.3446.

6. Prescott, M. K. M., Kashikawa, N., Dey, A. & Matsuda, Y. The Overdense Environment of a
Large Lyα Nebula at z ˜ 2.7. Astrophys J. Lett. 678, L77–L80 (2008). 0803.4230.

7. Saito, T. et al. Systematic Survey of Extended Lyα Sources over z ˜ 3-5. Astrophys J. 648,
54–66 (2006). arXiv:astro-ph/0605360.

8. Geach, J. E. et al. The Chandra Deep Protocluster Survey: Lyα Blobs are Powered by Heating,
Not Cooling. Astrophys J. 700, 1–9 (2009). 0904.0452.

9. Scarlata, C. et al. He II Emission in Lyα Nebulae: Active Galactic Nucleus or Cooling Radi-
ation? Astrophys J. 706, 1241–1252 (2009).

10. Colbert, J. W. et al. Polycyclic Aromatic Hydrocarbon Emission within Lyα Blobs. Astrophys
J. 728, 59–71 (2011).

11. Taniguchi, Y. & Shioya, Y. Superwind Model of Extended Lyα Emitters at High Redshift.
Astrophys J. Lett. 532, L13–L16 (2000). arXiv:astro-ph/0001522.

12. Mori, M., Umemura, M. & Ferrara, A. The Nature of Lyα Blobs: Supernova-dominated
Primordial Galaxies. Astrophys J. Lett. 613, L97–L100 (2004). arXiv:astro-ph/0408410.

13. Nilsson, K. K., Fynbo, J. P. U., Møller, P., Sommer-Larsen, J. & Ledoux, C. A Lyman-α
blob in the GOODS South ﬁeld: evidence for cold accretion onto a dark matter halo. Astron.
Astrophys. 452, L23–L26 (2006). arXiv:astro-ph/0512396.

14. Smith, D. J. B. & Jarvis, M. J. Evidence for cold accretion onto a massive galaxy at high
redshift? Mon. Not. R. Astron. Soc. 378, L49–L53 (2007). arXiv:astro-ph/0703522.

15. Fardal, M. A. et al. Cooling Radiation and the Lyα Luminosity of Forming Galaxies. Astro-
phys J. 562, 605–617 (2001). arXiv:astro-ph/0007205.

16. Yang, Y. et al. Probing Galaxy Formation with He II Cooling Lines. Astrophys J. 640, 539–552
(2006). arXiv:astro-ph/0509007.

7

17. Dijkstra, M. & Loeb, A. Lyα blobs as an observational signature of cold accretion streams
into galaxies. Mon. Not. R. Astron. Soc. 400, 1109–1120 (2009). 0902.2999.

18. Goerdt, T. et al. Gravity-driven Lyα blobs from cold streams into galaxies. Mon. Not. R.
Astron. Soc. 407, 613–631 (2010). 0911.5566.

19. Faucher-Gigu` re, C.-A., Kereˇ, D., Dijkstra, M., Hernquist, L. & Zaldarriaga, M. Lyα Cooling
e s
Emission from Galaxy Formation. Astrophys J. 725, 633–657 (2010). 1005.3041.

20. Lee, H. & Ahn, S. Polarization of the Ly alpha from an Anisotrophy Expanding H I Shell in
Primeval Galaxies. Astrophys J. Lett. 504, L61–L64 (1998). arXiv:astro-ph/9803085.

21. Dijkstra, M. & Loeb, A. The polarization of scattered Lyα radiation around high-redshift
galaxies. Mon. Not. R. Astron. Soc. 386, 492–504 (2008). 0711.2312.

22. Rybicki, G. B. & Loeb, A. Polarization of the LYalpha Halos around Sources before Cosmo-
logical Reionization. Astrophys J. Lett. 520, L79–L81 (1999). arXiv:astro-ph/9903291.

23. Prescott, M. K. M., Smith, P. S., Schmidt, G. D. & Dey, A. The Line Polarization within a
Giant Lyα Nebula. Astrophys J. Lett. 730, L25–L30 (2011). 1102.3918.

24. Stenﬂo, J. O. Resonance-line polarization. V - Quantum-mechanical interference between
states of diﬀerent total angular momentum. Astron. Astrophys. 84, 68–74 (1980).

25. Adams, T. F. The Escape of Resonance-Line Radiation from Extremely Opaque Media. As-
trophys J. 174, 439–448 (1972).

26. Neufeld, D. A. The transfer of resonance-line radiation in static astrophysical media. Astro-
phys J. 350, 216–241 (1990).

27. Kereˇ, D., Katz, N., Weinberg, D. H. & Dav´ , R. How do galaxies get their gas? Mon. Not. R.
s e
Astron. Soc. 363, 2–28 (2005). arXiv:astro-ph/0407095.

28. Dekel, A. & Birnboim, Y. Gravitational quenching in massive galaxies and clusters by clumpy
accretion. Mon. Not. R. Astron. Soc. 383, 119–138 (2008). 0707.1214.

29. Wardle, J. F. C. & Kronberg, P. P. The linear polarization of quasi-stellar radio sources at 3.71
and 11.1 centimeters. Astrophys J. 194, 249–255 (1974).

Acknowledgements Based on observations made with ESO Telescopes at the Paranal Observatory under
programme ID 084.A-0954. M.H. was supported in part by the Swiss National Science Foundation, and
also received support from Agence Nationale de la recherche bearing the reference ANR-09-BLAN-0234-
01. Patrick Ogle is thanked for his valuable suggestions about polarimetric observations. We are indebted
to Nino Panagia and Terry Jones for comments on the manuscript, and to Daniel Schaerer, Nick Scoville,
Chris Lidman, Arjun Dey, Moire Prescott, and Paul Lynam for discussions.

8

Author Contributions All authors contributed to the proposal preparation. M.H. and C.S. observed and
reduced the data. M.H. analyzed the results. All authors contributed to the manuscript preparation.

Competing Interests The authors declare that they have no competing ﬁnancial interests.

Correspondence Correspondence and requests for materials should be addressed to M.H.
(email: matthew.hayes@ast.obs-mip.fr).

9

Giant blob eso

More Related Content

Viewers also liked (9)

Similar to Giant blob eso (12)

More from Sérgio Sacani (20)

Recently uploaded (20)

Giant blob eso