Mehdi Lamee conducted a study using narrow band imaging to search for Lyman continuum (LyC) emission from Lyman alpha emitters (LAEs) at a redshift of ~3.3. He identified 178 LAE candidates in the Subaru/XMM-Newton Deep Survey field and stacked their LyC images to search for a detection, but found no detection. Correcting for IGM absorption, he placed an upper limit on the relative escape fraction of ionizing photons from LAEs of <0.27, corresponding to an absolute escape fraction of <0.21. The study aimed to constrain the escape fraction of LyC photons from high redshift galaxies and their potential role in reionizing the

1. Mehdi Lamee
Supervised by Dr. Claudia Scarlata
Preliminary Oral Examination
Minnesota Institute for Astrophysics
December 2013
NARROW BAND LyC IMAGING OF LAEs AT
Z~3.3 IN SXDS FIELD

2. OUTLINE
• Reionization and possible sources responsible for that
• The importance of LyC escape fraction
• LAEs
• Data
• Data reduction steps
• Looking for LAEs at z~3.3
• Stacking the LyC image
• Trying to estimate the relative escape fraction of LyC photons
• Discussing the sources of uncertainties

3. THE PROCESS OF REIONIZATION
• After cosmic dark ages something
should have ionized the universe!
• Reionization is done by z~ 6
• QSO could keep the universe
ionized at z < 2 but were not able
to contribute significantly to the
process of reionization.
Shaver et al. 1996

4. STAR FORMING GALAXIES AS CANDIDATES
Reionization Models predict young star forming
galaxies with reasonable LyC escape fractions
should be the main sources of reionization.
e. g. Robertson et al. 2013 claim:
Galaxies with MUV < -13 and LyC escape
fractions larger than 0.2 could ionize the
universe…
However, fesc is very uncertain.

5. LYMAN ALPHA EMITTERS
• LAEs are young star forming galaxies with strong Lyα emission line.
• LAEs on average have shown larger escape fraction.
• It seems the number of LAEs increases with redshift. (e.g Ouchi et al 2008)
• We can find them at high redshifts
• All of these are motivation to look for LAEs as possible analogs of galaxies
at the reionization epoch

6. WHY Fesc IS UNCERTAIN?
fesc
LyC
=
fobs
LyC
f LyC
int
• It is impossible to observe LyC beyond z > 4
• Must observe low redshift analogs
• At z < 3 space observation is need.
• Between 3 < z< 3.5 ground base telescope can
observe in UV
• But even from ground spectroscopy is very time
consuming.
• No need to mention these galaxies are faint!
In 2001 Steidel et al. introduced relative escape fraction: fesc,rel
LyC
= (
fobs
LyC
f LyC
int
)(
fobs
UV
f UV
int
)−1

7. NARROW BAND IMAGING OF LyC
• Inoue et al. 2005 used it for the first time.
• Much more efficient than spectroscopy for detecting larger sample
• Unlike spectroscopy it also probes wavelength much shorter than Lyman limit which suffer
more from IGM absorption.

8. IN THIS WORK…
• We used narrow band imaging technique
• Found LAEs at z~3.3
• Tried to constrain their LyC radiation
• Put upper limits on the relative escape fraction of ionizing photons

9. DATA
Subaru/ XMM Deep Survey, SXDS field. RA ~ 34.54 and DEC ~ -5.36
• 10 years of archival Suprime Camera data in U, B, IB527 and V bands.
• R band reduced image from Furusawa et al. 2008,
• Field of view: 34’ ×27’
Palomar/LFC observation, Sep 2011
• Central SXDS field
• New intermediate band filter IB383, samples rest frame 860Å < λ <910Å
• Field of view < 23’× 23’

10. DATA REDUCTION STEPS
Raw data
Combining all
exposures & making
the mosaic image

11. MEASURING FWHM OF THE IMAGE
• Measured averaged FWHM of
all exposures
• Removed the ones with bad
seeing
• Combined them
• Measured the FWHM and depth
of the mosaics

12. MAKING FINAL MOSAIC IMAGE
• Making exposure maps for each image:
• Averaging and combining:
• Making weight map:
(
flat
σi
)2

13. IB527 MOSAIC
5 arcmin

14. DEPTH OF THE MOSAICS
• Randomly distributing 10000 Apertures
with diameters of [1, 1.5, 2]FWHM
• Did the photometry
• Background was fixed to zero
• Fit a Gaussian profile to the left side of
histograms
• Right tail is due to objects
• Got the sigma of Gaussian

15. OBJECT DETECTION
• Running Sextractor in both dual and
single image modes
• IB527 as a detection image
• We masked out the area around bright
stars
• We calibrated the photometry using
CFHTLS catalog.

16. PHOTOMETRIC CALIBRATION
• CGHTLS T0006 as a reference catalog
• Objects with 18 < g & r <22 and 14 < u<
23
• SExtractor in single image mode, ZP=0
• If detected with >5σ and no flag!
• Matching tolerance < 0.3”
• Both catalog use SExtraactor mag_auto
• Choosing side-by-side bands
• Fit a line
• Remove outliers ( 2σ)
• Fit another line
• IBobs+ ZP = Ibcal
• IBcal – r =α(g – r)
• If needed BC03 galaxy SED templates
were used to get the ZP more precisely.

17. LYMAN ALPHA EMITTERS DETECTION
Lyα excess technique:
• Imaging with a combination of a narrow
and 1-2 broad band filters
• Lyα line excess of flux in NB
• Broad bands sample mostly the
continuum
• NB - BB color can be used to select LAE
candidates
Several people have used this technique to
discover hundreds of LAE candidates.
In our case, IB527 was used as a NB filter.

18. USING IB527 AND V FILTERS TO SELECT LAE
For an object with emission line
We define the observed line flux density in each filter to be: in erg/s/cm2/Å
Thus the observed flux density in each band can be written as
While we assumed fc is constant!
The equivalent width of the assumed line can be approximated to:

19. USING IB527 AND V FILTERS TO SELECT LAE
Doing some algebra we get:
Where
We also calculated the error bars on our approximated EW.

20. SELECTING 803 LAE CANDIDATES

21. EXCLUDING U BAND 3σ DETECTIONS
Objects with average V magnitudes
of 26.5 would have U magnitude of
~ 29
Our U band has 3σ detection limit of
27.1
Most of 140 U band detected
objects are low redshift interlopers.

22. EXCLUDING U BAND 3σ DETECTIONS

23. EXCLUDING 468 OBJECTS WITH B-V < 0.48
Due to stronger IGM
absorption in B band
We applied a cut at B – V=0.5
(Ouchi et al. 2008)
By excluding these objects we
make sure not to select
galaxies with extremely blue
SEDs instead of LAEs!

24. VISUAL INSPECTION

25. 178 OBJECTS PASSED OUR FILTERS!

26. LAE EXAMPLE
IB527LyC
VU
B
R

27. EXAMPLE

28. AVERAGE PROPERTIES OF OUR LAEs

29. LOOKING FOR LyC BY STACKING
• Making postage stamps of LAEs in LyC image
• Inspected visually and removed 62 LAEs
• Objects from the remaining 116 stamps were removed using V band segmentation map:
• Calculated the background and subtracted it from the image
• Stacking:
• Stacking images and do the photometry
• Do the photometry and stack fluxes
• S/N increases by √116

30. STACKING RESULTS
• NO detection!
• We still can put lower limits on the LyC magnitude
• We put lower limits on the FUV/FLyC

31. IGM ABSORPTION SIMULATION
• We followed the recipe of Inoue et al.
2008
• Absorbers distribution in any los:
• We start from z=0 and used the
distribution functions to assign Nhi, b
• We continues until the source redshift
z-3.34
• We calculated the optical depth of
each absorber in all frequencies…
Image credit: 2004 Edward L. Wright

32. IGM ABSORPTION SIMULATION
• We added all the optical depth to get the total optical depth for a los.
• Simulation covered 700Å <λ 1300Å with 1Å resolution.
• 4500 line of sights
• Taking the average:

33. APPLYING IGM CORRECTION
• On average flux in LyC band is decreased by 60% due to clumpy IGM!
• Correcting for that we get.

34. RELATIVE ESCAPE FRACTION
Defining
The relative escape fraction of LyC photons is:
But what is the value of ηint?
f LyC
esc,rel =
f LyC
esc
fUV
esc
=
ηint
η
exp(τIGM,LyC )

35. RELATIVE ESCAPE FRACTION…
Nestor et al. 2013

36. RELATIVE ESCAPE FRACTION…
Adopting a moderate value for ηint=3 we got:

37. LyC ESCAPE FRACTION?
Escape fraction of non-ionizing UV photons is also very uncertain!
Adopting a value of 0.3 for (Ouchi et al 2008, Blanc et al. 2011 )
We got
f LyC
esc,rel =
f LyC
esc
fUV
esc
fesc
UV
fesc
LyC
< 0.27
fesc
LyC
< 0.21

38. THANK YOU

39. ASTROMETRY
Reference catalog: CFHT-LS with
internal astrometric error of 0.02”
U band astrometry as an example.

40. MEASURING FWHM OF THE IMAGE
• Choosing objects from half-light-radii Vs
Mag plot
• Fitting a Gaussian profile horizontally &
vertically.
• Derive the ratio r = FWHM v / FWHM h
• Only use objects with 0.95< r < 1.05 for
averaging

41. MAKING FINAL MOSAIC IMAGE
• Making exposure maps for each image:
• Averaging and combining:
• Making weight map:
(
flat
σi
)2

42. DEPTH OF THE MOSAICS
• Randomly distributing 10000 Apertures
with diameters of [1, 1.5, 2]FWHM
• Did the photometry
• Background was fixed to zero
• Fit a Gaussian profile to the left side of
histograms
• Right tail is due to objects
• Got the sigma of Gaussian

43. PHOTOMETRIC CALIBRATION
• CGHTLS T0006 as a reference catalog
• Objects with 18 < g & r <22 and 14 < u<
23
• SExtractor in single image mode, ZP=0
• If detected with >5σ and no flag!
• Matching tolerance < 0.3”
• Both catalog use SExtraactor mag_auto
• Choosing side-by-side bands
• Fit a line
• Remove outliers ( 2σ)
• Fit another line
• IBobs+ ZP = Ibcal
• IBcal – r =α(g – r)
• If needed BC03 galaxy SED templates
were used to get the ZP more precisely.

44. MAGNITUDE HISTOGRAMS

45. USING IB527 AND V FILTERS TO SELECT LAE
Doing some algebra we get:
Where
We also calculated the error bars on our approximated EW.

46. DERIVATION OF EW

47. INOUE ET AL. 2008 ADOPTED DISTRIBUTIONS

48. INOUE ET AL. 2008 ADOPTED DISTRIBUTIONS

49. STOCHASTIC BEHAVIOR OF IGM

50. IGM ABSORPTION SIMULATION
• We followed the recipe of Inoue et al.
2008
• Absorbers distribution in any los:
• We start from z=0 and used the
distribution functions to assign Nhi, b
• We continues until the source redshift
z-3.34
• We calculated the optical depth of
each absorber in all frequencies…
Image credit: 2004 Edward L. Wright

51. EXAMPLE
IB527LyC
VU
B
R

52. WHY Fesc IS UNCERTAIN?
fesc
LyC
=
fobs
LyC
f LyC
int
• It is impossible to observe LyC beyond z > 4
• Must observe low redshift analogs
• At z < 3 space observation is need.
• Between 3 < z< 3.5 ground base telescope can
observe in UV
• But even from ground spectroscopy is very time
consuming.
• No need to mention these galaxies are faint!
In 2001 Steidel et al. introduced relative escape fraction: fesc,rel
LyC
= (
fobs
LyC
f LyC
int
)(
fobs
UV
f UV
int
)−1

53. LyC ESCAPE FRACTION?
Escape fraction of non-ionizing UV photons is also very uncertain!
Adopting three values of 0.2, 0.3 and 1 for we get:
f LyC
esc,rel =
f LyC
esc
fUV
esc
fesc
UV
fesc
LyC
< 0.18
fesc
LyC
< 0.14
fesc
LyC
< 0.27
fesc
LyC
< 0.21
fesc
LyC
< 0.9
fesc
LyC
< 0.7