The IR background as seen by WFC3
N.
1
Pirzkal ,
G.
1
Brammer
and the WFC3 team
1STScI
We present a new and improved characterization of the sky background light as seen by WFC3 in the the near-infrared. We
determine an empirical model of the zodiacal background from thousands of images obtained using WFC3 since its installation on
board of HST. While the structure of the background, as parametrized as a function of sun angle and ecliptic latitude, is similar to
the model contained in the WFC3 ETC, this new model extends the model to smaller values of sun angle with an accuracy of ~0.1
e-/s/pix for F098M, F105W, F125W, F140W, and F160W. We also present a characterization of the Earth-glow background as a
function of the HST pointing orientation with respect to the Earth limb. Interestingly, we identify a strong emission line component
of the background from metastable helium at 10,830 Å in the upper atmosphere that can significantly increase the background in
both IR grisms and in broad-band filters sensitive to this wavelength, even well above (> 40 deg) the bright Earth limb. Accounting
for these effects will be important for optimizing the efficiency of potential future deep integrations with the WFC3/IR grisms and
bluer broad-band filters.
WFC3 IR Zodiacal Light
Helium 10,830 Å Earth-glow
DATA
DIAGNOSIS
We measured the average background levels in individual IMA WFC3 exposures, first taking
care of masking out faint sources identified using SeXtractor. By keeping track of where HST
was pointing during each exposures, as parametrized by the Sun Angle and the Ecliptic
Latitude, we were able to obtain empirical measurements of the average Zodiacal light as
seen by HST using several commonly used broad and medium band filters. The filters we
considered and the number of images (IMSETs) suitable for this task are listed in Table 1.
The average backgrounds (in e-/s/pixel) measured in each filter, normalized to the level
seen in the F160W filter, are also shown. While measuring the Zodiacal light background,
only exposures not affected by the Bright Earth light were considered. Examples of
background levels as a function of Sun Angle and Ecliptic Latitude for the F125W and the
F160W filters are shown in Figure 1.
Variations in the background within an orbit timescale can be seen in some of the WFC3/IR
broad-band filters and grisms. A number of recent published works have noted strong
temporal variations within single exposures in a variety of WFC3/IR filters and grisms (Coe
et al. 2013; Koekemoer et al. 2013; Newman et al. 2013). The calwf3 pipeline assumes
constant count rates throughout an exposure in order to flag cosmic rays and bad pixels;
some of the observed background variations are strong enough so that a large fraction of
the entire detector is flagged as cosmic rays by calwf3, resulting in significantly nongaussian noise of the flt pipeline products.
–3–
F105W
Number of
FLTs
Number of
IMSETs
Scaling
Factor
F 098M
F 105W
F 110W
F 125W
F 140W
F 160W
297
704
636
1291
401
2852
485
1417
1283
1950
561
4411
0.68 ± 0.10
1.24 ± 0.20
2.05 ± 0.37
1.17 ± 0.11
1.48 ± 0.14
1.00
Table Table
1: Number
of individual
FLT included
in this in
ISRthis
forISR
eachforofeach
the filter
considered.
We also list
1: Number
of individual
FLT included
of thewe
filter
we considered.
the number of IMSETs (readout) in the associated IMA files. We measured the background levels in
We also
list the number
of IMSETs
(readout)
in the
associated
IMA
files.
We measured
individual
extensions
(i.e. readouts)
of these
IMA files
with
a sample
time
greater
than 100the
seconds.
background
in individual
(i.e.ofreadouts)
of these
IMAinfiles
with a sample
The last
column oflevels
this Table
list the extensions
scaling factor
back- ground
levels,
e−/s/pixel
observed in
timewith
greater
thanto100
The last column of this Table list the scaling factor of backthat filter
respect
theseconds.
F160W background.
Background (e- / s)
Filter
Jitter: LimbAng
100
q
The measured F125W (left) and F160W
(right) IR backgrounds in e /s/pixel, plotted as a
B( Sun
) = Angle.
1.5 ⇥ ( The
(1 +
( /1.5)2light1)background increases rapidly
(3) at small
function of Ecliptic Latitude and
Zodiacal
Sun We
Angle
values.
The shape
of the shown
observed
background
levels, as version
seen using
different
broad band
Fitting
fitted
the light
distribution
1 to simplified
of the
functional
e( ,in)Figure
= arccos(c(
, ))
(4)
filters, differs only by a scaling constant.
form introduced by Buffington et al. We found this representation to be a good match
Weoffitted
the
light
distribution
shown
Figure
1All
to simplified
version
of theoffunctional
of the
thebest-fit
Zodiacal
light
using inWFC3.
ofinthe
data
available
F098M,
The
values
forwe
thesee
ai parameters
are shown
Table
2. The
RMS for
the
fit is
form13%.
introduced
by Buffington
etand
al. F160W
We found
this
representation
towere
be anormalized
good match
F105W,
F110W,
F125W,
F140W
were
combined
after
they
to
MODELING
of the
of the
Zodiacal
lightvalues
we see
using
WFC3.
All of of
theTable
data1.available
for F098M,
the
F160W
levels
using
the
listed
in
the
last
column
The
combined
data,
We fitted the
light
distribution
shown
in Figure
1 one
to tothe
functional
form
introduced
by
The
values
listed
in
Table
2
and
Equation
1
allow
predict
the
expected
Zodiacal
F105W,
F110W,
F125W,
F140W
and F160W
were
combined
after
they
were
normalized
to
Buffington
et
al.
We
found
this
representation
to
be
a
good
match
of
the
of
the
Zodiacal
light
using
all
of
the
normalized
data
acquired
using
all
the
filters
listed
in
Table
1,
were
fitted
to
IR background in the F160W filter. Background levels in other filters can be determined by
the using
F160WWFC3.
levels using
values
in the last
column ofF105W,
Table 1. F110W,
The combined
data,
we see
All ofthe
the
datalisted
available
for F098M,
F125W,
F140W and
the
function
using
all of
the normalized
using all theto
filters
in Table
were fitted
to
F160W
were
combined
after data
theyacquired
were normalized
the listed
F160W
levels1,using
the values
listed
3
function
in thethe
last
column of Table 1. The combined data were fitted to the function
Z = a0 + a1 (1
where
Z = a0 + a1 (1
2
3
sin (B( ))
a6 (e( , )+a7 )
(1)
2
3
sin (B( ))
a6 (e( , )+a7 )
(1)
cos b( )) + (a2 + a3 c( , ) + a4 c( , ) + a5 c( , ) ) ⇥ 10
cos b( )) + (a2 + a3 c( , ) + a4 c( , ) + a5 c( , ) ) ⇥ 10
c( , ) = cos( ) ⇥ cos( )
where
q
2
B( ) = 1.5
⇥
(
(1
+
(
/1.5)
c( , ) = cos( ) ⇥ cos( 1))
e( , q) = arccos(c(2 , ))
B( ) = 1.5 ⇥ ( (1 + ( /1.5)
1)
(2)
(3)
(2)
(4)
(3)
e( , ) = are
arccos(c(
The best-fit values for the ai parameters
shown ,in ))Table 2. The RMS of the fit(4)
is
13%.
The data
best-fitlisted
values in
for the
ai parameters
shown in1Table
The RMS
the fit is best-fit
Fitting the
Table
1 using are
Equation
we 2.derive
the offollowing
listedofinthe
Table
Equation
1 allow This
one toconfirms
predict thethe
expected
parameters.
The RMS
fit 2isand
13%
on average.
valuesZodiacal
previously used
13%.The values
IRETC
background
F160Wthe
filter.
Background
levels in
other
filters
can of
be 50
determined
by
by the
as wellinasthe
extend
model
to the lower
Sun
Angle
limit
degrees.
The values listed in Table 2 and Equation 1 allow one to predict the expected Zodiacal
IR background in the F160W filter. Background
3 levels in other filters can be determined by
3
G102
G141
60
40
20
LimbAng = 20
LimbAng > 40
SHADOW
BrightLimb = 1
Predicted zodi
3.0
2.5
2.0
1.5
1.0
0.5
0.0
ibp329iqq
0
ibp329isq
10 20 30 40 50
t (minutes)
ib5x19tmq
0
ib5x19tqq
10 20 30 40 50
t (minutes)
ib5x21htq
0
ib5x21hwq
10 20 30 40 50
t (minutes)
ibkn06dmq
ibkn06dtq
ibkn06dhq
ibkn06dpq
0
10 20 30 40 50
t (minutes)
ibhj03xoq
0
ibhj03xvq
10 20 30 40 50
t (minutes)
3: Top:
Variation
of Earth limb
orbitthrough
for demonstrative
The curves
Fig. Figure
1.— Top
panels:
Variation
of angle
Earththrough
limb an
angle
an orbitexposures.
for demonstrative
are colored following the labels in the legend. The limb angle (“LimbAng”) and limb illumination
exposures.
The curves
arefrom
colored
following
the
in the condition
legend. is
The
limb angle
(“Lim(“BrightLimb”)
are taken
the time
sequence
andlabels
the SHADOW
estimated
from the
jit
files.and
Bottom:
of the (“BrightLimb”)
background count are
rate taken
throughout
orbit
for WFC3/IR
bAng”)
limb Variation
illumination
fromanthe
time
sequencefilters
andand
the
grisms, taken as the median pixel value of the central 400×400 pixels of each individual read in the
SHADOW
condition
is estimated
fromlevel
thefrom
shadoent
and model
shadoext
keywords
of theline.
jit
raw files.
The predicted
background
the zodiacal
is shown
in the dotted
F105W,
G102 and
G141 spectral
can count
show strong
increases in an
theorbit
background
files. The
Bottom
panels:
Variation
of the elements
background
rate throughout
for nueven > 40◦ above a dark Earth limb.
merous WFC3/IR filters and grisms, taken as the median pixel value of the central 400⇥400
pixels of each individual read in the raw files. The predicted background level from the
The dark “IR Blobs” seen in WFC3/IR images and the imaging flat-fields can be used to
Synphot/ETC
zodiacal model
shownthat
in the
dotted
line.
The F105W,
G102
and
G141in
estimate
the spectrum
of theissource
causes
the
variable
background
levels
shown
spectral
show strong
in the
> 40 shadow,
above
Figure
3. elements
In grism can
observations
takenincreases
during while
the background
telescope is ineven
the Earth
the
blobs
create
a smooth depression in the grism background corresponding to the
a dark
Earth
limb.
relativly smooth zodiacal continuum spectrum (Figure 4).
During times of elevated
background levels, however, the spectrum becomes point-like, suggesting a strong
emission line component. Extracting a negative spectrum of the IR blobs we find that the
2. Data
line is seen at the same wavelength in both the G102 and G141 grisms at 10,830 Å, which
we identify as coming from metastable helium atoms in the upper atmosphere and which
In order
to study
exposures
that
significant
portion
of an
orbit,
we obtain
has been
previously
observed
from
thesample
groundaduring
twilight
(Shefov
1961;
Bishop
& Link
from the MAST archive all public visits2 with exposure times longer than 600 sec for the
1999).
F105W filter. For the more commonly used F125W filter, we limit the search to exposures
longer than 1200
G141 grisms,
we obtain all of
the public
full-frame
Flatsec.
F140WFor the G102 andG141/Flat,
low background
G141/Flat,
high background
exposures
of non-crowded fields3 . These
searches
exposures inEmission
F105W,
a)
b)
Zodiacal result
continuumin 1083 c)
line 948 in
450
F125W,
1125 in G102, and 1754 in G141.
y
Fitting
Figure 1:
F160W
80
Filter observed
Number in
of that
Number
of respect
Scaling
ground levels, in e /s/pixel
filter with
to the F160W background.
FLTs
IMSETs
Factor
Filter Number of Number of
Scaling
F
098M
297
485
0.68
± 0.10
Fitting
FLTs
IMSETs
Factor
F 105W
704
1417
1.24 ± 0.20
F 110W
098M
297shown in Figure
485 1 to 2.05
0.68
± 0.37
0.10
We fitted the light
distribution
simplified
version of the functional
F
636
1283
±
F 125W
105W
704
1417this representation
1.24 ±
± 0.11
0.20 to be a good match
form introduced byF
Buffington et
al. We found
1291
1950
1.17
of the of the Zodiacal
light we 636
see using WFC3.
of the
data available for F098M,
F 140W
110W
1283
2.05
± 0.14
0.37
F
401
561 All 1.48
±
F105W, F110W, F125W,
and F160W were
after
they were normalized to
F 160W
125WF140W2852
1291
1950 combined
1.171.00
±
0.11
F
4411
the F160W levels using
the values401
listed in the 561
last column
of ±
Table
F 140W
1.48
0.141. The combined data,
using all of the normalized
using4411
all the filters listed
F 160W data acquired
2852
1.00 in Table 1, were fitted to
Tablethe
1: function
Number of individual FLT included in this ISR for each of the filter we considered.
We also list the number of IMSETs (readout) in the associated IMA files. We measured the
Table 1: Number
included
in this
ISR forofeach
the filter
considered.
background
levelsofinindividual
individualFLT
extensions
(i.e.
readouts)
theseof IMA
files we
with
a sample
sin (B( ))
2
3files. We
We
also
list
the
number
of
IMSETs
(readout)
in
the
associated
IMA
measured
the
, )+a7 )
6 (e(
Z
=
a
+
a
(1
cos
b(
))
+
(a
+
a
c(
,
)
+
a
c(
,
)
+
a
c(
,
)
⇥ 10 afactor
(1)
0
1
2
3
4
5
time greater than 100 seconds. The last column of this Table list the )scaling
of backbackground
levels
in/s/pixel
individual
extensions
(i.e.
readouts)
of these
IMAF160W
files with
a sample
ground
levels,
in
e
observed
in
that
filter
with
respect
to
the
background.
where
time greater than 100 seconds. The last column of this Table list the scaling factor of background levels, in e /s/pixel observed in
filter
respect
c( that
, )=
cos(with
) ⇥ cos(
) to the F160W
(2)
- background.
F125W
For every exposure in a defined association (some single orphan exposures that met the
search criteria but are not assigned an association are ignored), we measure the median flux
of the central
pixels400
of each of 250
nsamp300reads350
from the
tracks400
like
250 400⇥400
300
350
400 raw files
250to generate
300
350
x
x (l !)
x (l !)
those plotted in Fig. 1. The counts of read n is simply
the di↵erence of the [sci,
n] [sci, n+1]
Figure of
4: the
The raw
dispersed
grism spectrum
of the
takena during
extensions
files, which
we convert
to e“IR/sblobs”
assuming
single periods
gain ofof2.5elevated
e /DN
background levels (panel c) indicates that the excess background is dominated by an emission line at
for all10,830
amplifiers
and neglecting bias drift, linearity and flat-field corrections. The errors of
Å.
400
We find that the 10,830 Å line contributes to the IR background whenever the telescope
2
as is
of Jan.
6,
2014
itself
out
of
the Earth shadow. Its strength increases with decreasing target-to-limb
3
Program
11359, 11600,
12461,
11696,illuminated
12283, 12568, 12902,
13352, 13517,
12203, 12177,
angle,
i.e., IDs
a longer
path 12099,
length
through
atmosphere,
and 11597,
it contributes
an
12328, 11648,0.5-1
12927, e12547,
12471,
12330, 12970.
-/s to12190,
additional
the IR
background
even at large limb angles with the telescope
looking closer to zenith. For the worst case of day-side observations that graze the limb,
the observed background can be as much as 5-6 times higher than the nominal zodiacal
light for individual reads.
MITIGATION
ai
value
a0
a1
a2
a3
a4
a5
a6
a7
0.251549
0.0122746
0.652009
0.896383
1.24514
0.638263
0.00592507
136.589
the ai parameters in Equation 1
Observers planning background-limited observations in the Y-band at 1 µm are
encouraged to use the F098M filter in place of F105W, as the former is not sensitive to the
10,830 Å feature.
For archival exposures with strongly variable backgrounds that result in many pixels
flagged as cosmic rays and/or non-gaussian noise in the flt images, one can rerun
calwf3 setting CRCORR=OMIT in the raw files to turn off the linear-ramp fitting and
cosmic ray rejection. This technique loses some dynamic range for saturated sources
whose flux would normally be recovered from the unsaturated part of the ramp.
Furthermore, the flt products will contain all of the incident cosmic rays, though these can
be effectively identified given a sufficient number of dithered exposures with software such
as AstroDrizzle.
Fig. 2.— Our model of the F160W IR background in e /s/pixel as a function of Sun Angle