Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.mrao.cam.ac.uk/projects/OAS/publications/fulltext/spie4839-129-letter.ps Äàòà èçìåíåíèÿ: Wed Feb 20 17:36:25 2013 Äàòà èíäåêñèðîâàíèÿ: Sat Mar 1 03:29:01 2014 Êîäèðîâêà: IBM-866 Ïîèñêîâûå ñëîâà: mercury surface

Di#ractioní limited I band imaging with faint reference stars
Robert N.
Tu bbs a John E. Baldwin b and Craig D. Mackay a
a
Institu te of Astronomy, Cambridge University, Cambridge CB3 0HA, UK
b Cavendish Astrophysics
Grou p, Cambridge University, Cambridge CB3 0HE, UK
ABSTRACT
The use of
fain t
referen ce stars for the
selection of good short exposure images has
recen tly
been demon strated 1, 2
as a
techn ique which
can provide
essen tially
di#raction ílimited I
ban d
imagin g from wellífigured
groun díbased
telescopes as large as 2.5 m diameter. The
fain t
limitin g
magn itude
an d
en han ced
isoplan atic patch size for
the selected exposures
techn ique
mean s that 20% of
then ight sky is
within ran ge of a suitable
referen ce star
for
Iíban d
imagin g. Typically the 1%í10% of exposures with the highest Strehl ratios are selected.
When these
exposures are shifted
an d added together, field stars
in the
resultin g images have Strehl ratios as high as 0.26
an d FWHM as small as 90
milliarcsecon ds.
Within the selected exposures the
isoplan atic patch is
foun d to be
up to 50
arcsecon ds
in diameter at
810n m
wavelen gth. Images
within globular clusters
an d of multiple stars
from the Nordic Optical Telescope
usin g
referen ce stars as
fain t as I # 16 are
presen ted. The
techn ique relies
on an ew
gen eration of CCDs which provide
subíelectron readoutn oise at very fast readout rates. The
performan ce
of the
selection techn ique for various
astron omical programs is discussed
in comparison withn atural guide star
Adaptive Optics (AO).
Keyw ords: lucky exposures sky coverage
di#raction limited visible CCD
imagin g adaptive optics
1. INTRODUCTION
In recen t
observation s 1, 2 with the Nordic Optical Telescope we
demon strated that
selection of the best images
from a large dataset of short exposures
can provide
essen tially
di#raction ílimited images at
wavelen gths shorter
than 1 ²m
usin g
groun díbased telescopes as large as 2.5 m. The
techn ique was
shown to work reliably
usin g
referen ce stars as
fain t as I = 15.9,
an d
even 30
arcsecon ds away from the
referen ce star the FWHM of the image
poin tíspread
fun ction was as small as 130
milliarcsecon ds (without the use of
an y
decon volution techn iques).
The short exposure images were
taken with
an L3Vision CCD detector developed by E2V
Techn ologies
(formerly
Marcon i). These lowílight level CCDs allow fast frame rate
imagin g with low
readoutn oise. 3 It
is clear that they have
sign ifican t
sign al
ton oise
advan tages over both
con ven tion al CCD systems
an d imageí
in ten sified
photon ícoun tin g detectors. At low
sign al levels
photon ícoun tin g 4 with high
quan tum
e#cien cy is
possible
even at fast readout rates.
This report describes the method we have used for frame
selection ,
shiftin g
an d
coíaddin g,
an d
presen ts
some
prelimin ary results from the
techn ique. All the images
presen ted here were created from relatively short
observation s (2í10
min utes), but the
poten tial of the
techn ique to provide deep
di#raction ílimited images is
clear.
Further author
in formation :
(Sen d
correspon den ce to R.N.T.)
R.N.T.: Eímail:
rn t20@cam.ac.uk,
Telephon e: +44 1223 337296
J.E.B.: Eímail: jeb@mrao.cam.ac.uk,
Telephon e: +44 1223 337299
C.D.M.: Eímail: cdm@ast.cam.ac.uk,
Telephon e: +44 1223 337543

2. BACKGROUND
Momen ts
ofn earíperfect
seein g at large optical telescopes have
been discussed
in the literature
man y times over
a period of at least 60 years.
On e of the first detailed theoretical
an alyses of the
frequen cy of these ``lucky''
in stan ces of
excellen t image quality was published by Fried
in 1978. 5 He predicted that the probability of
an exposure
bein g
essen tially
di#raction ílimited (with Strehl ratio > 0.37) at a telescope of aperture diameter D
for
seein g
defin ed by Fried's 6 parameter r 0 would be:
P # 5.6 exp
#
-0.1557 (D/r 0 ) 2
#
This probability is a
stron gly
decreasin g
fun ction of aperture diameter, so that care must be
taken in choosin g
an appropriate aperture size. Hecquet
an d
Coupin ot 7 showed that apertures of 4r 0 - 7r 0 diameter are expected
to provide the highest
imagin g
resolution when selectin g the best 1% of exposures. For I
ban d
observation s
un der typical
Jun e/July
seein g
con dition s
con dition s8 the 2.56 m diameter of the Nordic Optical Telescope
correspon ds to the upper
en d of this
ran ge,
providin g a good compromise
between resolution an d
sen sitivity.
Ideally the exposure times used should be short
en ough to freeze
an y atmospheric
variation s. Exposure
times used for solar
imagin g are
often su#cien tly short,
an d image
selection techn iques have
been used very
successfully
in this field for
man y years. High frameírate CCD cameras have
recen tly provided Mercury imí
ages 9, 10 with suitable exposure times, but
observation s of
fain ter
astron omical targets have
been limited to
lon ger exposures due to
sign alítoín oise
con strain ts. The image quality provided by exposure
selection techí
n iques gradually degrades as exposure times are
in creased
un til there is little
improvemen t over local
seein g
con dition s. Selected exposures with
duration s of 0.1sí4s have
been very widely used for the
imagin g of
fain ter
Solar System objects, but
on on ly a
smalln umber of
occasion s has image
selection been used for objects outside
the Solar System
(n otable examples of image
selection observation s
in galactic
an d extraígalactic
astron omy
can be
seen in Refs: 11--16). Most of these studies have used image
cen troids for
reícen trin g the selected exposures
--n umerical
simulation s
in dicate that
reícen trin g exposures based
upon the
location of the brightest speckle
gives a
substan tial
improvemen t
in the
resultin g
resolution .
3. STREHL SELECTION TECHNIQUE
The image
selection techn ique we have
chosen for the
an alyses
presen ted here utilises Strehl ratios calculated
directly from images of
un resolved
referen ce stars
within the field of view. The applicability of this
techn ique to
fain t
astron omical targets
depen ds
on the probability that a
su#cien tly bright
referen ce star
can be
foun d
within the
isoplan atic patch
prevailin g at the times of the selected exposures. The
in ten sity of the brightest speckle
in the
referen ce star image provides a measure of the image Strehl ratio,
an d the
location of this speckle provides
the tipítilt
correction for the shift
an d add procedure. The probability of
fin din g a
suitablen atural
referen ce
starn ear
an astron omical target is expected to be very much higher for the selected exposures
techn ique
than it is
forn atural guide star AO for
an umber of
reason s:
1. The processes of image
selection an d image
reícen trin g require a similar
amoun t of light to that of tipítilt
correction (in the selected exposures a large
fraction of the starlight falls
in on e speckle,
improvin g the
sign al
ton oise for tipítilt
measuremen ts). With our selected exposures
techn ique the full light
collectin g
power of a # 7r 0 telescope aperture is used, whereas AO
wavefron t
sen sors must perform tipítilt
correction usin g a subíaperture of
aroun d 1r 0 diameter.
2. AO systems use a feedback loop to correct
wavefron t errors, so the exposure times required
in AO
wavefron t
sen sors are typically as short as 0.1t 0 . The selective exposures
techn ique is passive so that
an exposure
time of t 0 is adequate (where t 0 is the atmospheric
coheren ce time).
3. At the times
when the selected exposures are acquired, the
wavefron ts
en terin g the telescope are flat to
within #1
radian rms after tipítilt
correction .
Usin g the
relation ship
between r 0
an d the
varian ce of the
wavefron t phase across
an aperture after tipítilt
correction 17 it is possible to calculate
an in stan tan eous
value of r 0 appropriate for the selected exposures, which will have a value
substan tially larger
than

the
mean r 0 for the local
seein g
con dition s. The improved
wavefron t
flatn ess is expected to provide a
substan tial
improvemen t
in the
isoplan atic patch size for the
techn ique as compared to
con ven tial speckle
imagin g
techn iques. For the best 0.3% of exposures selected at a telescope of diameter 7r 0 the
isoplan atic
patch should be doubled
in size. The
en han cemen t to the
isoplan atic patch size
mayn ot be as
pron oun ced
at rare
in stan ces
when improved
wavefron t
flatn ess is brought about by two phase
screen s widely separated
in the atmosphere
can cellin g each other out.
Poin ts 1.
an d 2. above should theoretically lead to
an improvemen t of
aroun d 6
magn itudes
in the
limitin g
magn itude of
referen ce star which
can be used for image
selection , compared to
then atural guide star AO case.
The
fraction of exposures to be selected is decided
durin g post
processin g of the data. The
selection criteria
used
determin es the
sen sitivity, image
resolution an d
isoplan atic patch size,
an d
can be tailored to
then eeds of
the
in vestigator.
4. LOW LIGHT LEVEL CCD (L3VISION)
In man y previous high frame rate
imagin g
observation s, detector
readíoutn oise has compromised the image
quality
an d
substan tially reduced the
limitin g
magn itude which
can be
obtain ed. The
recen t
developmen t of
CCDs
withn egligible
readoutn oise 3 has
elimin ated
then oise
pen alty for fast readout,
providin g
an improvemen t
in the
limitin g
magn itude of
referen ce star which
can be used
an d
allowin g the
detection of
fain t objects
in the
field.
The
techn ology
behin d the
L3Vision detectors is disclosed
in Ref: 18.
In essen ce a
con ven tion al CCD
structure is used with the output register
exten ded with
an addition al
section that has
on e of the three phases
clocked with a much higher voltage
than isn eeded purely for charge
tran sfer. The large electric fields that are
established
in the
semicon ductor material
ben eath pairs of serial
tran sfer electrodes cause charge carriers to
be accelerated to high velocities.
Addition al charge carriers
can then be
gen erated by impact
ion isation . The
charge
multiplication per
tran sfer is really quite small, typically
on e
percen t but with a
largen umber of
tran sfers
(591 for the device used here)
substan tial
electron ic
gain s may be achieved. The output of this
exten ded serial
register is passed
on to a
con ven tion al CCD output amplifier. The
electron icn oise of this amplifier
isn ow
divided by the
gain factor of the
multiplication register which, if this
gain is high
en ough, will reduce the
e#ective output
readn oise to levels much smaller
than on e
electron rms. The
gain register
in troduces
addition al
n oise to the device
domin ated by the statistics of the
amplification process.
Thisn oise is
proportion al to the
squareíroot of the
sign al level. At low
sign al levels
an d high
multiplication register
gain the
readoutn oise is
n egligible
an d
photon ícoun tin g
can be performed.
5. OBSERVATIONS AND DATA REDUCTION
On 2001 July 6th we
un dertook high frame rate
imagin g
observation s at the
Cassegrain focus of the 2.56 m
Nordic Optical Telescope
usin g a camera built
aroun d a
lowín oise
L3Vision CCD. The camera comprised a
fron tíillumin ated 576 ½ 288 pixel
frameítran sfer CCD65 with 20 ½ 30 ²m pixels, cooled
in a
liquidn itrogen dewar to
min imise the dark
curren t. The CCD was
run by
an AstroCam 4100
con troller modified to provide a
variable voltage clock
sign al for the output
gain register of the CCD. We used frame rates
between 18 Hz
an d
140 Hz, with subíarray readout
wheren ecessary.
The f/11 beam at the focus was
con verted to f/60
usin g a
sin gle achromat,
givin g
an image scale of 27 ½
40
milliarcsecon ds per pixel
an d a total
imagin g area of 11.5 ½ 15.4
arcsecon ds.
In order to
in vestigate the
image quality achieved at large
an gular
separation s from the
referen ce star, the camera optics were
design ed so
that light from two
region s of the sky separated by 25 arcsec could be superimposed
on the CCD. This allowed
scien ce targets to be imaged at
separation s of up to 30
arcsecon ds from the
referen ce star. All the
observation s
were made at
810n m with a topíhat filter of
120n m
ban dwidth
an d
withn o autoguider
in operation .
Addition al
observation s were
taken for two hours
on 2002 July 25th through severe
Saharan dust
extin ction at the Nordic Optical Telescope,
an d
again on 2002 July 26th through
sign ifican tly clearer skies. It
hasn ot
been possible to fully reduce
an y of the data from these
run s, but two
prelimin ary images based
on run s of

Figure 1. The
isoplan atic patch for the selected exposures
techn ique was
in vestigated
usin g
bin ary star systems with
a
ran ge of
an gular
separation s.
On e
bin ary
compon en t
was used as the
referen ce star for image
selection ,
shiftin g
an d
coíaddin g. The Strehl ratio of the
bin ary
compan ion divided by the Strehl ratio for the
referen ce star is plotted
again st the
an gular
separation between the
bin ary comí
pan ion s.
Figure 2.
Depen den ce of image
resolution on referen ce
star
magn itude for a
ran ge of
referen ce stars
in M13. Typí
ical image FWHM are
given in milliarcsec for stars
within a few arcsec of the
referen ce star, for a
ran ge of
referen ce
stars.
5í10
min utes
observin g time have
been in cluded (Figures 5
an d 6, discussed
in Section 6). The pixel scale for
these
observation s was 25 ½ 37
milliarcsecon ds.
In order to
con vert the short exposure images
taken on the
rectan gular pixel CCD array
in to square pixel
arrays
an d to improve the
samplin g of the images, the
in dividual short exposure images were resampled
usin g
sin c
in terpolation to give 4 times as
man y pixels
in thex coordin ate
an d 6 times as
man y pixels
in the y
coordin ate.
The accuracy of
measuremen t of the Strehl ratio
an d
position of the brightest speckle was improved by
con volvin g the
in dividual frames with the
di#raction ílimited telescope
poin tíspread
fun ction before these paí
rameters were calculated. The peak
in ten sity
in the
con volved images
represen ts the
location where
correlation with the
di#raction ílimited
poin tíspread
fun ction is maximised,
providin g a very good estimate for the
location an d
in ten sity of a
di#raction ílimited speckle
within the
origin al image. After frames had
been selected based
on the Strehl ratio for the brightest speckle
foun d, the
correspon din g
un con volved exposures were shifted
an d
coíadded to produce the
fin al image.
6. RESULTS
The size of the
isoplan atic patch
in Strehlíselected exposures was deduced from
observation s of the
bin ary stars
HD 203991, 8 Lac
an d 61
Cygn ii, which have
separation s of 0.6, 22
an d 30 arcsec respectively. These were all
bright
en ough that the
selection an d image shift
correction for exposures was free of the e#ects of
readoutn oise
an d
smalln umber statistics
in photon detection .
On e
compon en t of each
bin ary was used as the
referen ce to
select the best 1% of the exposures, for which both the image full width at halfímaximum
in ten sity (FWHM)
an d Strehl ratio of the other
compon en t were measured. Figure 1 shows the
fraction al
reduction in Strehl ratio
of the
secon d
compon en t as a
fun ction of
an gular
separation from the
referen ce star. A
Gaussian fit to the
poin ts gives 50
arcsecon ds for the FWHM of the
isoplan atic patch for these selected frames,
taken durin g a
period
in which the
mean seein g was 0.51 arcsec. The image FWHM for objects 30 arcsec from the
referen ce
star
usin g the best 1% from a
run of 4000 exposures of 61
Cygn ii was 130 milliarcsec,
on ly slightly poorer
than the FWHM of # 80 milliarcsec expected for a well sampled
di#raction ílimited
poin tíspread
fun ction . The
measured FWHM is
in creased to 230 milliarcsec
when selectin g the best 10% of exposures,
an d 300 milliarcsec
when selectin g all the exposures for a
shiftían díadd image.
The e#ect of
referen ce star
magn itude
on the quality of the
resultin g image was studied
usin g
observation s
with the full field of view of the CCD
in a
den sely populated
region of M13
un der 0.46
arcsecon d
Iíban d

seein g
con dition s
(correspon din g to 0.51
arcsecon d
seein g at
500n m). The frame rate for these
observation s
was limited to 18 Hz,
allowin g image
motion to slightly blur the exposures. This limited the Strehl ratios for
recon structed images to # 0.16, with image FWHM of # 100 milliarcsec. Data was
taken from 6
run s of 1000
frames each. The
selection of the best 1% of the 6000 exposures worked successfully with
referen ce stars as
fain t as I = 15.9. The e#ects of
``overíresolution '' 16
on fain t
referen ce stars were avoided by
usin g other objects
in the field for image quality
measuremen ts. Figure 2 shows the
variation in the image FWHM
ofn earby stars
when a
ran ge of
di#eren t stars are used as the
referen ce for image
selection ,
shiftin g
an d
addin g. I
ban d stellar
magn itudes were
taken from Ref. 19. The Strehl ratios with
referen ce stars of I = 13.8
an d I = 15.9 are
0.13
an d 0.065 respectively, a
substan tial
improvemen t over the value of # 0.019 for
con ven tion al
astron omical
images
gen erated by
summin g all of the exposures
in a
run without
recen terin g the exposures.
It is clear from Figure 2 that the
measuremen t of the Strehl ratio
an d
position of the brightest speckle works
well with
referen ce stars as
fain t as I = 15.9
usin g our
existin g CCD camera.
When combin ed with the large
isoplan atic patch size
demon strated by Figure 1, this allows
imagin g over a very
substan tial
fraction of the
n ight sky -- 8% of the South Galactic Pole
region (based
on 400 stars per square degree 20 with I # 15.9)
an d
an average of 20% over the whole sky.
In comparison , I
ban dn atural guide star adaptive optics is typically limited
to I # 10 for highíorder
correction , 21 with a
sign ifican tly smaller
isoplan atic patch
an d
hen ce very much poorer
sky coverage
aroun d suitable
referen ce stars.
Figure 3a shows a small field
in M13
aroun d the bright star labelled A
in Ref. 19.
An I = 12.7
referen ce
star was used for image
selection an d
reícen trin g as
shown in the Figure. Slight asymmetry
in the
fain t halos
aroun d the stars may have resulted from poor charge
tran sfer
e#cien cy at low
sign al levels
on the CCD -- a
result of the
un usually low
operatin g temperature. For this figure the best 10% of exposures were selected,
givin g typical stellar FWHM of # 120
milliarcsecon ds. This compares very favourably with the FWHM
on HST WFPC2 images -- typically 140
milliarcsecon ds
on the PC chip
an d 190
milliarcsecon ds
on the Wide Field
CCDs. Cross
section s through I = 13.8
an d I = 14.9 stars are
shown in Figure 3b.
The best 1% of exposures selected from this 5.5
min ute
observation of M13
usin g the same
referen ce star
gave
an image with FWHM of < 0.1
arcsecon ds
an d showed stars as
fain t I = 19. For
an hour of
observin g this
limitin g
magn itude will be
in creased to at least I = 21.
In order to asses the image quality, the 60 best exposures of M13 were separated
in to two groups of 30. The
exposures
in each group were shifted
an d added together
givin g two
in depen den t images of the field
in M13.
These images were
then con volved with the
di#raction ílimited telescope
poin tíspread
fun ction .
Measuremen ts
were made of the
location an d
in ten sity of the brightest pixel for each star,
providin g crude relative astrometry
an d photometry
within the images. The rms
di#eren ce
in astrometry
between the two
in depen den t datasets was
foun d to be 6 milliarcsec,
an d the rms
di#eren ce
in stellar
magn itudes was 0.02. With
lon ger
observation s the
accuracy of the astrometry
an d photometry is expected to improve
substan tially,
poten tially
allowin g accurate
measuremen ts of globular cluster velocity
dispersion s
an d
groun díbased photometric variability studies.
Figure 4 shows a
fieldn ear the
cen tre of the globular cluster M15. The I = 13.1 star circled was used for
exposure
selection ,
shiftin g
an d
coíaddin g. This image was
gen erated by
selectin g the best 1% of exposures
from 2000 frames
taken at 18 Hz frame rate.
Figure 5 shows
an image of
an earby field
in M15 which has
been con trastíen han ced to highlight some of
the
fain test stars. The
location of the cluster
cen tre as
determin ed
in Ref: 22 is marked with a cross. Over
70 stars brighter
than I = 19 are visible
in this 13 ½ 10
arcsecon d
region alon e. This image was
gen erated
from the best 3% of exposures
taken durin g 5
min utes of
observin g time at the NOT. The image FWHM are
typically 240½140
milliarcsecon ds --
althoughn ot
di#raction ílimited, this
represen ts a
substan tial
improvemen t
over the 0.6
arcsecon d
con ven tion al
(seein gílimited) image. The
reduction in resolution when compared with
observation s of M13
in July 2001 is largely attributable to the very
stron g
depen den ce of the image FWHM
in selected exposures
on the local
seein g
con dition s. For the
observation s of M13 the
lon g exposure image FWHM
an d short exposure image Strehl ratios are
con sistan t with
an Iíban d r 0 size of 0.36 m, with the telescope
aperture
hen ce
measurin g 7.1r 0 diameter (this
Iíban d r 0 size would
correspon d to a
500n m
lon gíexposure
seein g disk of 0.51
arcsecon ds FWHM). For the
observation s of M15, the r 0 size was 0.28 m,
makin g the
aperture diameter equal to 9.2r 0 . The image
resolution obtain ed from
selectin g a few
percen t of exposures is

Figure3 . The best 10% of exposures from a field
in M13 were selected, shifted
an d added to produce image a. The
referen ce star for frame
selection an d
recen trin g has
been labelled. North is at the top.
Crossísection s through the
I = 13.8
an d I = 14.9 stars labelled X
an d Y are
shown in part b, with FWHM of # 120
milliarcsecon ds.
kn own to drop rapidly with aperture diameter for apertures greater
than # 7r 0 . 7 With bright
referen ce stars
an improvemen t
in resolution can be
obtain ed by
stoppin g
down the telescope aperture, at the
expen se of a loss
of
sen sitivity.
The 5
min ute
observation of the
region shown in Figure 5 is the
on ly data which has so far
been an alysed
from
then ight of 2002 July 26th. Eight
adjoin in g fields were also observed
on thatn ight
an d we
even tually
in ten d to form a mosaic from all of the data.
Figure 6 shows
an image of the multiple star system Gliese 569,
gen erated by
selectin g the best 4% of
exposures from 10
min utes of
observin g time
on 2002 July 25th. The
fain t
compan ion visible
in the
in set image
is thought to be a triple
brown dwarf system. 23 The widest
separation in the triple system is slightly larger
than the
di#raction limit of the NOT at
810n m,
an d
un der more typical
Jun e/July
seein g
con dition s it should
be possible to estimate
an Iíban d
magn itude
di#eren ce for the two
prin iciple
compon en ts.
Two short exposures
taken from a
run on a
fain t star are
shown in Figure 7. Figure 7a shows a surface
plot of the
in ten sity
in a selected exposure
(on e with high Strehl ratio), while Figure 7b shows the exposure
with the
median Strehl from the same
run . The
improvemen t
in sign al
ton oise for the
shiftían díadd process
provided by exposure
selection is
apparen t from the
di#eren ce
between the peak
in ten sity
in 7a
an d 7b. It is
this
sign al
ton oise
improvemen t allows the image
selection techn ique to use
fain ter
referen ce stars
than those
used for
con ven tion al
shiftían díadd.
7. FUTURE PROSPECTS OF EXPOSURE SELECTION
The exposure
selection techn ique described here has the
poten tial to provide
di#raction ílimited
imagin g from
an y medium sized
groun díbased telescope which has wellífigured optics
an d good
astron omical
seein g
con dition s.
For
highíresolution observation s the
techn ique requires a telescope aperture which
isn o more
than 7r 0 diameter.

2 arcsecond
Figure 4. The best 1% of exposures from a
fieldn ear
the
cen tre of M15 were selected shifted
an d added to
produce this image. The I=13.1
referen ce star used for
frame
selection an d
recen trin g has
been circled. The exí
posures were selected from data
taken durin g 2
min utes
of
observin g time at the NOT. North is to the lower left.
Figure 5.
Con trastíen han ced image of a 13 ½ 10
arcsecon d
region aroun d the core of M15. This image was
gen erated
from the best 3% of exposures
taken durin g 5
min utes of obí
servin g time at the NOT. The
location of the cluster
cen tre
is marked with a cross.
If the
seein g
con dition s are 0.5
arcsecon ds at
500n m,
then 7r 0
correspon ds to a diameter of 2.9 m at
900n m
an d
8.5 m at 2.2 ²m.
Seein g
con dition s are
kn own to vary
durin g
then ight, 24
an d if
seein g degrades
substan tially
then high
resolution observation s
can on ly be performed after the telescope aperture has
been stopped
down .
The
magn itude limit of this image
selection techn ique will be improved by the
developmen t of
thin n ed
an tií
reflection coated
L3Vision CCDs. With such a device
on e could expect to detect I # 23
un resolved sources
in an hour of
observin g if the device was operated
in photon coun tin g mode at a 2.5 m telescope. This would allow
scien tifically
rewardin g studies such as the accurate
measuremen t of velocity
dispersion s
in globular clusters,
an d the
poten tial for accurate
groun díbased photometry
in crowded fields such as globular clusters
an d of
smallíseparation gravition al
len ses.
The rarity of ``Lucky Exposures''
mean s that image
selection techn iques are rather wasteful of starlight,
makin g them
un popular with
man y telescope operators. However, the ability to perform
Iíban d
di#raction í
limited
imagin g of fields of up to 50
arcsecon ds diameter
usin g
referen ce stars so
fain t that they are widely
available across the
wholen ight sky has
en ormous
scien tific
poten tial for mediumísi ed
groun díbased telescopes.
Natural guide star
Iíban d adaptive optics systems are
on ly capable of
operatin g
in the
vicin ity of relatively bright
referen ce stars,
an d
can n ot compete with exposure
selection for
imagin g the vast majority of
astron omical
sources.
The relatively low cost
an d ease of
con struction of high frame rate
L3Vision cameras make them attractive
proposition s for a
largen umber of mediumísi ed optical telescopes. The probability of
obtain in g a ``lucky
exposure'' is
substan tially
in creased if the lowíorder
wavefron t
aberation s are corrected,
makin g this image
selection techn ique applicable to larger telescopes which have AO systems optimised for
wavefron t
correction at
in fraíred
wavelen gths
(hen ce
providin g lowíorder
correction in the visible). Further
discussion of shortíexposure
image
selection with AO systems
can be
foun d
in Refs: 25--27.

0.2``
Figure 6. The best 4% of exposures of the multiple star
system GJ 569. The
fain t object
on the right is thought to
be a
brown dwarf triple system. 23 The images were selected
from 10
min utes of
observin g time at the NOT,
on an ight
with high
Saharan dust
extin ction at the observatory
an d
slightly belowíaverage
seein g
con dition s for July.
Figure 7. Surface plots of
in ten sity
in two example expoí
sures
taken from a
run on a
fain t star. Plot a) shows
an exposure with high Strehl ratio, while plot b) shows the
exposure with the
median Strehl ratio from the same
run .
The
improvemen t
in sign al
ton oise for the
shiftían díadd
process provided by exposure
selection is
apparen t from
the
di#eren ce
between the peak
in ten sity
in a)
an d b).
8. CONCLUSIONS
The exposure
selection techn ique described here has the
poten tial to provide
di#raction ílimited
imagin g from
an y medium sized
groun díbased telescope which has wellífigured optics
an d good
astron omical
seein g
con dition s.
The
techn ique works well over a larger
isoplan atic patch
than is accessible to adaptive optics,
an d
can utilise
referen ce stars which are much
fain ter
than those required for adaptive optics.
With
Iíban d
observation s at the 2.56 m Nordic Optical Telescope we have
demon strated that the image
selection techn ique
can providen ear
di#raction ílimited
Iíban d
imagin g over 20% of
then ight sky
usin g low
light level detectors
recen tly developed by E2V
Techn ologies. Fields as large as 50
arcsecon ds
in diameter
aroun d a
referen ce star
can be imaged at
highíresolution ,
an d the use of
referen ce stars as
fain t as I # 15.9
has
been demon strated. The
developmen t of
thin n ed
an tiíreflection coated
L3Vision CCDs will improve the
referen ce star
magn itude limit to
fain ter
than I = 16
an d allow the
detection of sources
in the field as
fain t as
I # 23
in an hour of
observation .
ACKNOWLEDGMENTS
The Nordic Optical Telescope is operated
on the
islan d of La Palma
join tly by
Den mark,
Fin lan d,
Icelan d,
Norway
an d
Sweden ,
in the
Span ish Observatorio del Roque de los Muchachos of the
In stituto de AstrofÒÐsica
de
Can arias.

This project has
been supported by the
European Commission through the Access to
Research Infrastructures
Action of the Improving Human Potential Programme, awarded to the
In stituto de AstrofÒÐsica de
Can arias to
fun d
European Astron omers' access to the
European Northern Observatory,
in the
Can ary
Islan ds.
REFERENCES
1. R. N. Tubbs, J. E.
Baldwin , C. D. Mackay,
an d G. C. Cox,
``Di#raction ílimited CCD
imagin g with
fain t
referen ce stars,'' A&A 387, pp. L21--L24, May 2002.
2. J. E.
Baldwin , R. N. Tubbs, G. C. Cox, C. D. Mackay, R. W.
Wilson ,
an d M. I.
An dersen ,
``Di#raction í
limited
800n m
imagin g with the 2.56 m Nordic Optical Telescope,'' A&A 368, pp. L1--L4, Mar. 2001.
3. C. D. Mackay, R. N. Tubbs, R. Bell, D. J. Burt, P. Jerram,
an d I. Moody,
``Subelectron readn oise at MHz
pixel rates,''
in Proc. SPIE, 4306, pp. 289--298, May 2001.
4. P. Jerram, P. J. Pool, R. Bell, D. J. Burt, S.
Bowrin g, S.
Spen cer, M. Hazelwood, I. Moody, N. Catlett,
an d P. S. Heyes, ``The LLCCD: lowílight
imagin g without
then eed for
an in ten sifier,''
in Proc. SPIE, 4306,
pp. 178--186, May 2001.
5. D. L. Fried, ``Probability of
gettin g a lucky shortíexposure image through
turbulen ce,'' Optical Society of
America Journal 68, pp. 1651--1658, Dec. 1978.
6. D. L. Fried, ``Optical
Resolution Through a
Ran domly
In homogen eous Medium for Very
Lon g
an d Very
Short Exposures,'' Optical Society of America Journal 56, pp. 1372--1379, 1966.
7. J. Hecquet
an d G.
Coupin ot, ``A
gain in resolution by the
superposition of selected
recen tered short expoí
sures,'' Journal of Optics 16, pp. 21--26, Feb. 1985.
8. C.
Mun ozíTun on , J.
Vern in ,
an d A. M. Varela, ``Nightítime image quality at Roque de LOS Muchachos
Observatory,'' Astronomy and
Astroph ysics Supplement Series 125, pp. 183--193, Oct. 1997.
9. R. F.
Dan towitz, S. W. Teare,
an d M. J. Kozubal,
``Groun díbased
highíresolution imagin g of mercury,''
AJ 119, pp. 2455--2457, May 2000.
10. J.
Baumgardn er, M.
Men dillo,
an d J. K.
Wilson , ``A Digital
HighíDefin ition Imagin g System for Spectral
Studies of
Exten ded
Plan etary Atmospheres. I.
In itial Results
in White Light
Showin g Features
on the
Hemisphere of Mercury
Un imaged by
Marin er 10,'' AJ 119, pp. 2458--2464, May 2000.
11. J.íL. Nieto, A. Llebaria,
an d S. di Serego Alighieri,
``Photon ícoun tin g detectors
in timeíresolved
imagin g
mode í Image
recen trin g
an d
selection algorithms,'' A&A 178, pp. 301--306, May 1987.
12. J.íL. Nieto, S. Roques, A. Llebaria, C.
Van derriest, G. Lelievre, S. di Serego Alighieri, F. D. Macchetto,
an d M. A. C.
Perryman ,
``Highíresolution imagin g of the double QSO 2345 + 007 í
Eviden ce for subcomí
pon en ts,'' ApJ 325, pp. 644--650, Feb. 1988.
13. G. Lelievre, J.íL. Nieto, E.
Thouven ot, D.
Salmon ,
an d A. Llebaria, ``Very high
resolution imagin g
usin g
subípupil apertures,
recen terin g
an d
selection of short exposures,'' A&A 200, pp. 301--311, July 1988.
14. D.
Crampton , R. D. McClure, J. M. Fletcher,
an d J. B.
Hutchin gs, ``A search for closely spaced
gravitation al
len ses,'' Astronomical Journal 98, pp. 1188--1194, Oct. 1989.
15. J.íL. Nieto, M. Auriere, J. Sebag, J.
Arn aud, G. Lelievre, A. Blazit, R. Foy, S.
Bon aldo,
an d E.
Thouven ot,
``The optical
coun terpart of the Xíray
bin ary
in the globular cluster NGC 6712,'' A&A 239, pp. 155--162,
Nov. 1990.
16. J.íL. Nieto
an d E.
Thouven ot,
``Recen trin g
an d
selection of shortíexposure images with
photon ícoun tin g
detectors. I í Reliability tests,'' A&A 241, pp. 663--672,
Jan . 1991.
17. R. J. Noll,
``Zern ike
polyn omials
an d atmospheric
turbulen ce,'' Optical Society of America Journal 66,
pp. 207--211, Mar. 1976.
18. D. Burt
an d R. Bell, ``CCD imagers with
multiplication register,'' European Patent Application Bulletin
39, p. EP 0 866 501 A1, Sept. 1998.
19. R. L.
Cohen , P. Guhathakurta, B.
Yan n y, D. P.
Schn eider,
an d J. N. Bahcall, ``Globular Cluster Photometry
with the Hubble Space Telescope.VI.WF/PCíI
Observation s of the Stellar
Population s
in the Core of M13
(NGC 6205),'' AJ 113, pp. 669--681, Feb. 1997.
20. N. Reid
an d G. Gilmore, ``New light
on fain t stars. II í A photometric study of the low
lumin osity
main sequen ce,'' MNRAS 201, pp. 73--94, Oct. 1982.

21. J. E. Graves, M. J. Northcott, F. J. Roddier, C. A. Roddier,
an d L. M. Close, ``First light for hokupa'a:
36íelemen t curvature ao system at uh,'' SPIE Proceedings 3353, pp. 34--43, Sept. 1998.
22. P. Guhathakurta, B.
Yan n y, D. P.
Schn eider,
an d J. N. Bahcall, ``Globular Cluster Photometry With the
Hubble Space Telescope. V. WFPC Study of M15's
Cen tral
den sity Cusp,'' Astronomical Journal 111,
pp. 267--+,
Jan . 1996.
23. M.
Ken worthy, K.
Hofman n , L. Close, P.
Hin z, E. Mamajek, D. Schertl, G. Weigelt, R.
An gel, Y. Y. Balega,
J.
Hin z,
an d G. Rieke, ``Gliese 569B: A
Youn g Multiple
Brown Dwarf System?,'' ApJL 554, pp. L67--LL70,
Jun e 2001.
24. R.
Racin e, ``Temporal
Fluctuation s of Atmospheric
Seein g,'' Publications of
th e Astronomical Society of
th e Pacific 108, pp. 372--+, Apr. 1996.
25. C. D. Mackay, J. E.
Baldwin ,
an d R. N. Tubbs, ``Noise free detectors
in the visible
an d
in frared:
Implication s
for the
design ofn extígen eration ao systems
an d large telescopes,''
in Proc. SPIE, 4840, p.
(in press), Aug.
2002.
26. M. C.
Roggeman n , C. A. Stoudt,
an d B. M. Welsh, ``Imageíspectrum
sign alítoín oiseíratio
improvemen ts
by statistical frame
selection for adaptiveíoptics
imagin g through atmospheric
turbulen ce,'' Optical Engií
neering 33, pp. 3254--3264, Oct. 1994.
27. S. D. Ford, M. C.
Roggeman n ,
an d B. M. Welsh, ``Frame
selection performan ce limits for statistical image
recon struction of adaptive optics
compen sated images,''
in Proc. SPIE Vol. 2534, p. 235í246, Adaptive
Optical Systems and Applications, Robert K. Tyson; Robert Q. Fugate; Eds., 2534, pp. 235--246, Aug.
1995.