Asteroid color photometry with Gaia
and synergies with other space missions
Marco Delbo
UNS-CNRS-Observatoire de la Côte d’Azur
&
Gaia DPAC Coordination Unit 4 (Solar System Objects)
May 23, 2011
Pisa, Italy
Collaborators
Philippe Bendojya (UNS-CNRS-Observatoire de la Côte d’Azur)
Antony Brown (Leiden Observatory, the Netherlands)
Giorgia Busso (Leiden Observatory, the Netherlands)
Alberto Cellino (INAF-Osservatorio Astronomico di Torino, Italy)
Laurent Galluccio (UNS-CNRS-Observatoire de la Côte d’Azur)
Julie Gayon-Markt (UNS-CNRS-Observatoire de la Côte d’Azur)
Christophe Ordenovic (UNS-CNRS-Observatoire de la Côte d’Azur)
Paola Sartoretti (Observatoire de Paris)
Paolo Tanga (UNS-CNRS-Observatoire de la Côte d’Azur)
Outline of the presentation
Introduction
Asteroid spectral classes and mineralogy
Modern CCD based asteroid spectroscopy and its limitations
Asteroid spectral classification using Gaia
The BP-RP photometers on board of Gaia
Expected peformances of the BP-RP
Data products for asteroid color photometry
Asteroid spectral classification algorithm
Unsupervised clustering algorithm for asteroid spectral
classification
Combination of Gaia photometry with AUXILIARY Data
Asteroid spectral types
Asteroids are assigned a type based on spectral shape. These types
are thought to correspond to an asteroid’s surface composition.
Bus and Binzel spectral types:
I
1.2
C-group (carbonaceus) with a
featureless spectrum
S
1.1
1.05
I
I
B-type (featureless and blue)
S-group (stony) with silicate
absorption bands
S-type
X-type
B-type
C-type
1.15
Reflec
tance
X
C
1
B
0.95
0.9
0.85
I
X-group of mostly metallic
objects including
enstatite-chondrite like spectra
0.8
0.4
0.45
0.5
0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
Wavelength (microns)
from the SMASS web site, by R. Binzel and collaborators
o
re
ra
n
n
y
tof
y
is
d
o
he
l.
ut
d
d
bn
cm
e
o
m
d,
n
e
p
e
g
o
asteroids described by Tholen.
In spectral component space, the C-complex is not centrall
condensed, but rather a bifurcated cloud with two broad, rela
Fig. 13. view of the C-complex in PC4! versus PC1! space. Here we see the C-types plotting
tively
distinct
concentrations of points that are best separate
PC4! values plot DeMeo
further to the right.et al., 2009
in the spectral component plane described by PC3! and PC2
as seen in Fig. 9. This bifurcation results from two population
spe
within the C-complex that are differentiated based on the pre
tax
ence or absence of a broad absorption feature, centered nea
ing
0.0
0.7 µm. In addition to the 0.7-µm feature, the component plan
dat
in Fig. 9 is sensitive to the strength of the UV absorption shor
est
ward of 0.55 µm. The order in which the C-complex was divide
wav
into taxonomic classes was based on the dominance of theseal.,
tw
wit
spectral features in component space. Spectra containing a dee
tion
UV absorption
feature were classified first, followed by objec
a v
Fig. 4. PC2! versus PC1! plotted for the C- and X-complexes plus D-, K-, L-, and T-types
slop
whose spectra include a 0.7-µm band. Finally, those spectra wit
shallow Fig.
to nonexistent
UV
features
were
subdivided,
based
pr
14. Prototype examples for C- and X-complex spectra.
to
witd
marily on their spectral slope.
mem
pon
ized
a pronounced
dropoff asteroids
similar to the
Cgh,the
butBdoes
notC-type
clas
eige
Inbyseparating
hisUVG-class
from
and
show the 0.7-µm feature that define Ch and Cgh. The classes Cg,
Tholen
had the advantage of colors derived from the ECASden
s
Cgh, Ch, Xk, Xc, and Xe do not all separate cleanly in component
Modern CCD based asteroid spectroscopy
Bus and Binzel 2002
FIG. 8. Examples of spectra contained in the S-, Sk-, Sq-, Sr, Sa, and
Sl-classes, presented in the same format as Fig. 3. Plotted are the spectra of
5 Astraea (S), 3 Juno (Sk), 33 Polyhymnia (Sq), 2956 Yeomans (Sr), 1667 Pels
(Sa), and 169 Zelia (Sl), showing the range of characteristics found within the
core of the S-complex. At the center of this core are the S-type asteroids, such
as 5 Astraea. The spectrum of Astraea contains a moderately steep UV slope
shortward of 0.7 µm (a), and a moderately deep 1-µm silicate absorption band
that exhibits a concave-up curvature, with a local minimum centered near 0.9 µm
(b). Trends in UV slope and 1-µm band depth noted in Fig. 5 are demonstrated
here, with the Sl-, Sa-, and Sr-class spectra containing steeper UV slopes than
the Sk- and Sq-type spectra. Similarly, the Sk- and Sl-class asteroids have the
shallowest 1-µm bands, while Sr-type asteroids exhibit the deepest 1-µm band
FIG. 10. Examples of spectra contained in the B-, Cb-, C-, Cg-, Ch-, and
depth. Over the range of the UV slope shortward of 0.7 µm, the interval from
Cgh-classes, presented in the same format as Fig. 3. Plotted are the spectra of
0.44 to 0.55I
µm is often slightly steeper than the interval from 0.55 to 0.7 µm,
142 Polana (B), 191 Kolga (Cb), 1 Ceres (C), 175 Andromache (Cg), 19 Fortuna
(Ch), and 706 Hirundo (Cgh), showing the range of spectral characteristics
found in theI
C-complex. The spectra of both 142 Polana and 191 Kolga are
essentially featureless and are differentiated only on the basis of their slope.
C-type asteroids, such as 1 Ceres, have spectra containing a weak to moderately
strong UV absorption shortward of 0.55 µm (a). The UV absorption in the
spectrum of 175 Andromache is considerably stronger (b), accounting for its
pon
PC1
space because their distinguishing features are weak and not well
PCi
PC1
detected by the first five principal components. These classes ofcon
tenseen
mustinbe
by visually
detecting
features described
disr
as
thedistinguished
spectra of both
3 Juno and
2956 Yeomans
(c). The spectra
PC1
by Bus (1999). A summary of these features are described at the
ber
both
3 Juno and 169 Zelia contain 1-µm bands that show moderate concave-u
end of the flowchart, Appendix B. Fig. 14 shows typical spectra for
The
PC1
curvature (d), differentiating these Sk- and Sl-class asteroids from the Kan
classes within the C- and X-complexes.
L-ty
L-classes, respectively. A subtle inflection in the UV slope of 33 Polyhymn
R-ty
T
is centered near 0.63 µm (e). This feature has been previously recognized
2 w
cha
4.6. A near-infrared-only classification method
the
combination of two absorption bands, centered at roughly 0.60 and 0.67
µ
cau
as
Fig. 5. Examples of S-, Sa-, and A-classes. There is a clear progression from S-types
(Hiroi
et al. 1996).
ver
V-cl
with a shallow 1-µm band and low slope to A-types with a deep 1-µm band and
For many objects, data exist in either the visible or nearbet
high slope. Sa- and A-types show similar 1-µm band absorptions, but Sa-types are
fam
infrared
wavelength
ranges
butandnot
such
much
less red
than A-types.
The class
the both.
asteroidWhile
numbertaxonomies
are labeled next
to
and
hav
each
spectrum.
as the
Bus system (Bus, 1999; Bus and Binzel, 2002b) are available
thre
nya
for visible data, no system has been widely accepted for assigning
X-c
surv
to data
existing
theor
near-infrared.
Weshallow
have adapted
aclasses
minimum
near
1 µm only
and in
may
may not have
2-µm
as n
mem
our presentband;
taxonomy
interpret
spectral
available
only in
absorption
it alsototends
to be
steeplydata
sloped.
The Sa-class
abl
clas
the the
near-infrared
range. This1-µm
adaptive
taxonomy
to
has
same characteristic
absorption
bandisasnot
themeant
A-class,
prin
1-µm
but
is less a
red.
determine
definite class, but instead is an intermediate tool to inthe
However, all spectra do not go shortwards 450nm.
Most available data in the blue region (340-550nm) are very
poor in quality.
Sloan Digital Sky Survey (SDSS); Parker et al. (08) colors
SDSS: color photometry of more than 100,000 asteroids.
Example from the SDSS Moving Object Catalog 4 (MOC4).
1.1
Eos family
colors ind
within a
ences betw
longing to
1
3.1. A meth
1.2
Reflectivity
Asteroid families in the SDSS MOC
S-type
Traditio
objects in
cluster de
analysis (
strong col
can be use
the mixin
tion.
The SD
color and
0.9
0.8
0.7
0.6
0.5
0.3
0.4
0.5
0.6
0.7
Wavelength (um)
0.8
0.9
1
a∗ ≡ 0.89(
bands: u’:354, g’:477, r’:623, i’:763, z’:913 (mn)
with a∗ = 0.89(g 0 − r 0 ) + 0.45(r 0 − i 0 ) − 0.57.
Fig. 3. A plot of the color distribution in a∗ and i − z of 45,087 unique objects listed
in both the SDSS MOC 4 and ASTORB file, and that have H corr < 16. The approximate
boundaries of three spectral classes are marked, and used in labeling family type.
The color-coding scheme defined here is used in Figs. 4–6.
The a∗ co
distributio
transform
et al., 200
formed by
bution of
principle c
and by N
are well
a∗ = 0.49
Asteroid spectral classes and mineralogy of the main belt
Fig. 3. A plot of the color distribution in a∗ and i − z of 45,087 unique objects listed
in both the SDSS MOC 4 and ASTORB file, and that have H corr < 16. The approximate
boundaries of three spectral classes are marked, and used in labeling family type.
The color-coding scheme defined here is used in Figs. 4–6.
I
Investigation of the
mineralogy of families.
I
Comparison of spectra of
NEAs with those of families
near the NEA source
regions...
with the help of dynamical
models; see e.g. De Leon et al.
2010; Campins et al. 2010;
Jenniskens et al. 2010; Walsh et
al. 2011; Gayon-Markt et al.
2011; etc..)
Fig. 4. A plot of the proper a vs. sin(i ) for the same objects as shown in Fig. 3. The
color of each dot is representative of the object’s color measured by SDSS, according
to the color scheme defined in Fig. 3. The three main regions of the belt, defined by
strong Kirkwood gaps, are marked.
the technique developed by Ivezić et al. (2002b) to visualize this
The photometers onFocal
the focal
plane of Gaia
plane
1 pixel 60 x 180 mas
106 CCDs (4.5 x 2 kpix) = 1 Gpixel
SM1-2
BP
AF1 - 9
RP
RVS
420 mm
0.69°
WFS
WFS
BAM
BAM
sec
FOV1
0s
10.6
15.5
30.1
49.5
56.3
64.1
5.8
10.7
25.3
44.7
51.5
59.3
FOV2
0s
sec
disclaimer: in the Gaia community, BP-RP data is called color phometry;
it is low resolution (R ∼ 20 − 90) slit-less spectroscopy, though.
A job for
Rosario
The photometers:
resolving
power (R =
R~20
R~70
R~90
R~70
λ
∆λ )
GaiaI will
take lowis such to have about 18
Sampling
resolution
spectra bands in the BP-RP
independent
of about
200000
domain (A. Brown, spring 2011)
asteroids
I Sampling is 60 pixel per
(SNR~50-100)
photometer, signal is in general
Taxonomy
contained in 40 pixel per
Space weathering
age ofphotometer.
the
I Telescope
asteroids
from PSF FWHM is about 2
Lecce/Catania
pixels AL (40/2∼20 independent
model?bands) and 1 pixel AC.
I
80% of asteroid observations have
velocities ≤15 mas/s.
Beacuse a CCD transit lasts 4 s →
≤1 pixel widening of the PSF: this
is not too bad.
PSFs vary with the Field of
View and different focal
plane position
calibration of this effect is called
geometrical calibration, both in the AL
direction (by means of a reference sample)
and in the AC direction (by means of 2D
windows);
BP-RP response for point like sources
d
window
focal pl
in the P
G=15 point source with different colors.
40
sample
30
20
10
0
60
40
sample
30
20
10
2.5
2.0
1.5
1.0
0.5
2.5
2.0
1.5
1.0
0.5
0.0
0.0
400
500
! (nm)
0
E
a se
3.0
RP counts (10 3 photons)
BP counts (10 3 photons)
3.0
50
680 900
640 700
800 9001000
! (nm)
The s
Fields
a
Non
pecte
G2V star is thecalibrated
middle green curve.
Internally
spectra
Credits: Busso, G. & Brown, A. 2009
External Calibration: the aim is to obtain externally calibrated s
BP-RP SNR for an asteroid with G=17
14
BP
RP
12
SNR
10
8
6
4
2
0
300
400
500
600
700
800
900
Wavelength (nm)
Photon Noise limited in general. So SNR=
p
Nphotons
1000 1100
BP-RP SNR as function of magnitude (1 transit)
SNR [400-1000] nm
1000
Min SNR
Peak SNR
100
10
1
10
12
14
16
G magnitude
18
20
I
Minimum and Peak SNR in the range 400-1000 nm per
transit.
I
Best fit to min SNR: SNR=17631 × 10−0.201317∗G
Average SNR for BP-RP at the end of mission
I
I
The large majority of asteroids
(main belt) are observed at least
60 times [Mignard, F. 2001
(SAGFM09)]
The SNR of the accumulated
(avarage spectrum) is 8 times
larger
SNR [400-1000] nm
10000
Min SNR
Peak SNR
1000
100
10
10
12
14
16
G magnitude
18
20
Minimum SNR at the end of the
mission assuming 50 transit/asteroid
For asteroids with G=19-20 spectral classification will be difficult.
Solution:?!: Spectral binning.
Spectral Shape Coefficients: 8-colors asteroid survey
160
140
I
I
Spectral Shape Coefficients (SSCs;
4 for BP and 4 for RP; 8 colors for
each source) are calculated by
IDT (Initial Data Treatment).
SSCs calculated also by PhotPipe
and refined at every cycle.
Potentially very interesting for
performing an 8-colors asteroid
survey.
Signal (e-/transit)
120
I
BP
RP
BP-SSC
RP-SSC
100
80
60
40
20
0
300 400 500 600 700 800 900 1000 1100
Wavelength (nm)
Example of SSC values calculated
for a BP-RP signal of a G=20
asteroid (solar-like spectrum).
Data products for asteroid color photometry
I
The spectral energy distribution (SED) is obtained from
accumulated BP-RP data.
I
I
I
I
Asteroid reflectivity is calculated from the SED.
I
I
I
Average SEDs is produced (1 per asteroid).
Epoch SEDs is also produced where possible (for SNR≥20 per
transit ∼G≤15).
Smearing due to proper motion is taken into account.
BP and RP SEDs are combined into one SED.
The SED is divided by the solar spectrum and the results
normalized at 0.55 microns.
The asteroid reflectivity is used to determine the asteroid
spectral class.
I
I
Unsupervised clustering algorithm.
Comparison with other classifications (e.g. Bus & Binzel).
Clustering method based on Minimal Spanning Tree (MST)
Example of MST in R2
!"#
Galluccio et al. (2008 )
Method for partitioning a set V of N data
points (V ∈ RL ) into K non-overlapping
clusters with:
I
the inter-cluster variance is maximized;
I
the intra-cluster variance is minimized.
!"!&#
!"$
!"%
!"!&
!"&
!"'
!"!'#
!
!!"'
!"!'
!!"&
!!"%
!"!!#
!!"$
!!"#
!!"#
Identification of the number of clusters:
!"#
!"#
!
!
Figure 1. Left: construction of a MST through a rando
!"$
I
!
!"!&#
a new edge built at iteration i.
The lenght of the edge at each additionof of
a vertex of the MST is recorded.
!"%
!"!&
!"&
!"'
I
!"!'#
Spectral clustering with a graph-
!
Then by identifying valleys in this curve,
In [3], Grishkat et al propose a new distance
we can estimate the number and positions
when two MST rooted at each vertex intersect.
of high density regions of points → i.e. from each initial vertex, until its collapse. This
Hitting Time. More recently, a new distance m
the clusters.
has been deducted from the previous. Details
!!"'
!"!'
!!"&
!!"%
!"!!#
!!"$
!!"#
!!"#
!
!"#
!
!
#!
'!!
'#!
&!!
&#!
%!!
%#!
$!!
of thisa random
distance,
tog(i)
as=Dual
Figure 1. Left: construction of a MST through
set ofreferred
data. Right:
|ei |, theRooted
length P
of a new edge built at iteration i.
We propose to exploit these similarity measure
Test of the classification algorithm
I
Spectra of asteroids belonging to all spectral classes were
obtained at the Telescopio Nazionale Galileo (TNG) under
Gaia-like observing geometry.
PI Paolo Tanga; Data analysis in progress.
photo credits: P. Tanga
I
See next talk by Julie Gayon-Markt.
Removal of spectral classification degeneracies
1.2
Normalized Reflectance
1
0.8
0.6
But asteroids (46) Hestia, (55)
Pandora, and (317) Roxane have
different albedos.
1.2
Reflectance Normalized to the albedo
There are some well known
degeneracies in the mineralogical
interpretation of asteroid spectral
classes.
For instance, asteroids (46)
Hestia, (55) Pandora, and (317)
Roxane have very similar spectra.
(55) Pandora - M-type
(46) Hestia - P-type
(317) Roxane - E-type
1
0.8
0.6
0.4
0.2
0
0.3
0.4
0.5
0.6
0.7
Wavelength (um)
0.8
0.9
0.4
0.2
(55) Pandora - M-type
(46) Hestia - P-type
(317) Roxane - E-type
0
0.3
0.4
0.5
0.6
0.7
Wavelength (um)
0.8
0.9
1
Albedo + spectra → removal of
spectral class degeneracies.
1
Asteroid spectral classification (Gaia + WISE data)
I
NASA WISE has observed 100,000 asteroids in the thermal IR.
I
Albedos will be obtained from WISE data.
I
First data (IR images) already released.
I
Albedo + spectra → removal of spectral class degeneracies.
I
Albedo and spectra can be classified using our non supervised
classification algorithm.
ExploreNEOs with Warm Spitzer: PI D. Trilling (NAU)
Albedos from Warm Spitzer
1.0
0.8
e
0.6
0.4
Albedo (%)
0.2
0
0.0
0.5
1.0
1.5
5
2.0
a (AU)
10 15 20 25 30
2.5
3.0
Conclusions
DPAC products (from Gaia observations only):
I
Gaia will obtain R ∼ 20 − 90 visible spectra of asteroids.
I
Average spectra (reflectancies) will be published.
I
Epoch spectra for the brighter asteroids.
I
Spectral classes of asteroids will be also published.
Gaia + Auxiliary data (e.g. WISE albedos):
I
Albedo from WISE or Spitzer will allow spectral classes
degeneracies to be removed
→ mineralogical map of the main belt.