Espectroscopía de alta resolución con el nuevo instrumento GRACES

Espectroscopía de alta resolución con
el nuevo instrumento GRACES
Eder Martioli
Investigador Científico y Gerente de la
Oficina Gemini Brasil (BrGO)
GRACES
15 años de ciência con Gemini - La Plata, 02 de Junio de 2015.
CONTENIDO
1. El instrumento GRACES
2. Los datos de GRACES
3. El proyecto OPERA y la reducción de los datos
4. Oportunidad cientifica con GRACES
GRACES
15 años de ciência con Gemini - La Plata, 02 de Junio de 2015.
GRACES
Gemini Remote Access to CFHT
ESPaDOnS Spectrograph
Gemini
CFHT
Source: A-N Chene et al. SPIE 2014 paper
\
ESPaDOnS
Echelle SpectroPolarimetric Device for the Observation of Stars at CFHT
ESPaDOnS
ESPaDOnS con GRACES!
ESPaDOnS con GRACES!
Los datos de ESPaDOnS y GRACES
flat
Th-Ar
objeto
fabry-perot
bias
polarimetria
ESPaDOnS vs. GRACES
etro
m
i
lar
Po
CFHT - ESPaDOnS
modos fibras
star-only:
polar:
∥ ⟂
star+sky:
ø1.6”
star
star
sky
ø1.6”
ø2.2”
10”
slices
Gemini - GRACES
modos
fibras slices
star-only:
star+sky:
ø1.2”
star
star
sky
ø1.2”
ø1.2”
1’
ESPaDOnS vs. GRACES
Configuración
instrumental
ESPaDOnS
GRACES
Modo
star
star+sky
star
star+sky
focal ratio
f/8
f/8
f/16
f/16
number of slices
6
3
4
2
number of fibers
1
2
1
2
fiber size (μm)
100
100
165
165
fiber length (m)
36
36
270
270
resolution
80,000
67,000
55,000
33,000
aperture size
(arcsec)
1.6
1.6
1.2
1.2
EL PERFIL INSTRUMENTAL
DE GRACES
Star+Sky Low-Res
Star-Only Hi-Res
EL PROYECTO OPERA (CFHT)
Open-source Pipeline for ESPaDOnS Reduction and Analysis
Web:
http://sourceforge.net/projects/opera-pipeline/
Paper: Software and Cyberinfrastructure for Astronomy II. Proceedings of the SPIE, Volume
8451, id. 84512B-84512B-21 (2012)
Paper: GRACES: Gemini remote access to CFHT ESPaDOnS spectrograph through the
longest astronomical fiber ever made: experimental phase completed, Chené, A.-N. et al.,
Proceedings of the SPIE, Volume 9151, id. 915147 16 pp. (2014)
Paper científico en preparación
Reducción de los datos con OPERA
1. Calibración
Datos de calibración -> para generar productos de calibración
2. Reducción
Productos de calibración -> para extraer los espectros de
ciencia
3. Análisis
Obtener informaciones de los datos reducidos
Reducción de los datos con OPERA
Calibración
1.
2.
3.
4.
5.
6.
Master combining
Gain and Noise
Geometry
Instrument Profile
Aperture
Wavelength calibration
Reducción
1.
2.
3.
4.
5.
6.
Optimal Extraction
Telluric Wavelength Correction
Heliocentric Correction
Flat-fielding
Normalization by continuum
Flux calibration
Reducción con OPERA “in a nutshell”
Ganancia/Ruido Geometría
Perfil Instrumental
Abertura
Perfil Instrumental
Abertura
Calibración en longitud
de onda
Extracción
Flat-fielding / Normalización / Calibración en flujo
operaInstrumentProfileCalibration
Perfil instrumental super amostrado bidimensional
como función de las coordenadas de la imagen
Flujo
Primero la medición del Perfil Espacial
abertura (a)
CFHT IP for Star-Only 80k - 6 slices / 1 fiber
0.07
0.06
0.06
0.05
0.05
0.04
0.04
0.03
IP
IP
CFHT IP for Polar 65K - 3 slices / 2 fibers
0.07
0.03
0.02
0.02
0.01
0.01
0
0
-0.01
0
5
10
15
20
25
30
-0.01
-20
35
-15
-10
-5
0
pixels
pixels
CFHT IP for Special Mode - 2 slices / 3 fibers
5
10
15
20
CFHT IP for S+S 65k - 3 slices / 2 fibers
0.06
0.08
0.07
0.05
0.06
0.04
0.05
0.03
IP
IP
0.04
0.03
0.02
0.02
0.01
0.01
0
0
-0.01
0
5
10
15
20
pixels
25
30
35
-0.01
0
5
10
15
20
pixels
25
30
35
Después viene la medición de un perfil
instrumental 2D super amostrado
orientación
líneas
1 píxel
abertura (a)
columnas
Perfil instrumental 2D super amostrado
Modo Star-only
Fabry-Perot
Th-Ar
operaExtractionApertureCalibration
La abertura para la extracción
Calibración de la abertura utilizando un perfil
instrumental bidimensional
masterincalibration
frame.
ADU
Fast readout
mode. Nominal gain and noise values for ESPaDOnS in Fast readout mode are 1.5 e−/ADU and
e−,and
respectively.
b. 4.7
Gain
Noise: measure CCD gain and noise using on a set of flat-field and bias exposures. As an example, for data
obtained on 2014-05-06 the measured gain was 1.68±0.02 e−/ADU and the noise was 5.51 e−, with a bias level of 490
ADU in Fast readout mode. Nominal gain and noise values for ESPaDOnS in Fast readout mode are 1.5 e−/ADU and
4.7 e−, respectively.
LOS 2 MODOS DE OPERACIÓN DE GRACES
Figure 8. Left panel shows a section of a flat-field frame in Star-Only (4-slice) mode, showing a few orders in the far red. Right
panel shows the same section of a flat-field frame in Star+Sky (2-slice) mode.
Figure 8. Left panel shows a section of a flat-field frame in Star-Only (4-slice) mode, showing a few orders in the far red. Right
panel shows the same section of a flat-field frame in Star+Sky (2-slice) mode.
Chene
et al. isSPIE
2014 paper
Figure 9. Top panels show measurements of the instrument profile for orders 35Source:
and 36,A-N
where
left panel
in Star-Only
mode
and right panel is Star+Sky mode. Bottom panels show the same measurements with aperture sub-pixels masked out.
1
Figure 9. Top panels show measurements of the instrument profile for orders 35 and 36, where left panel is in Star-Only mode
and right panel is Star+Sky mode. Bottom panels show the same measurements with aperture sub-pixels masked out.
IRAF (http://iraf.noao.edu) is distributed by the National Optical Astronomy Observatories, which are operated by the Association of Universities for
Research in Astronomy, Inc., under cooperative agreement with the National Science Foundation.
operaWavelengthCalibration
Calibración píxel / longitud de onda utilizando
g. 19.— Example of two apertures with heights of 0.6 (left panel) and 1.0 pixel (rig
un
espectro
de
Th-Ar
anel). The 2D IPs are plotted with the aperture sub-pixels masked out for better isuali
on.
1. Algoritmo para detectar líneas espectrales utilizando
2.7.entre
Wavelength
Calibration
correlación cruzada
la imagen
y el perfil bidimensional
The module operaWavelengthCalibration performs the pixel-to-wavelength calib
defrom
las alíneas
un atlas
de Th-Ar
on, where2.
it Identificación
detects spectral lines
Th-Ar con
comparison
exposure
and combine th
nes with known atlas data to create a calibration polynomial solution of the form:
3. Ajuste polinomial
(d) = a0 + a1 d + a2 d2 + a3 d3 + ... + anp dnp ,
(3
where4.
d is
the
distance
vector
given
by
equation
35,
a
0,...,np are the coefficients and np
Utiliza superposición de los órdenes para generar
e maximum order of the polynomial. For ESPaDOnS we found that it is sufficient to u
polynomial with np =una
3. solución única y homogénea
Algoritmo de
detección de
líneas
Medición de los
centros de las
líneas
algorithm is applied for an oversampled tilted aperture, in which th
from a two-dimensional instrument illumination profile as explained
of a 2D IP for flux extraction seems to be a necessary calibration for
operaExtraction
tilted profiles, typically
found in fiber-fed spectrographs that use
pseudo-slit and with limited CCD pixel size, such as in ESPaDOnS.
Extracción
del espectro
The three methods
for extraction
implemented in operaExtrac
1. RAW: Raw Sum;
2. STD: Standard Extraction (with background subtraction);
3. OPT: Optimal Extraction.
The algorithms for each of these methods are described in mor
3.1.1. Raw Sum
xc,yc
x
y
the sums must use or reject the same set of pixels.
Extracción
Raw Sum
Note that the pixel values (Ci00 ,j 0 ) 3.1.1.
of the master
comparison image are
– 38 –
image indexes (i, j), therefore we have to use the following relationship to c
This
method
obtainsindexes:
the energy flux density spectrum by countin
the
image
and sub-pixel
and events
Mi0 j 0 is ainbad
pixel
mask given
by:
photo-electron
each
spectral
element.
The flux is given by
i,j
1pixel
x, y
values over an extraction area defined
by the extraction aperture calibr
⇢
0
1 FLOOR(x
if of
Ii0 j 0both
<c +
SATURATION
& mi0 j 0 = 1a
i
=
x
0 ), detector readout
and the variance includesMcontributions
the
i
0
0
ij =
otherwise,+ y 0 ),
j 0= FLOOR(y
0
c is jgiven
The raw flux fi0 j 0 for each individual
sub-pixel
by:
where the quantity mi0 j 0 is a boolean for a pre-defined bad pixel mask
– 38 –0in the parenthesis do
0
where
the function
FLOOR()
rounds(i,the
number
easily obtain
the image
pixel indexes
j)
from
indexes
(i
(Ii0 j 0 the
Bi0sub-pixel
)GA
s
j0
closest
i0 j 0 = Mi0 j 0and yc is sampled in ,steps of 1 p
31 integer,
and 32. xc (y) is the tracefpolynomial
F0 0
andimage.
Mi0 j 0 is aEquations
bad pixel mask
given
the extension of the
26 and
27by:
give ixj0i0 and yj0 0 .
The output spectrum is given by an array of spectral elements, whe
0
2 of ⇢
Analogously
to the
defined
in
equation
letsub-pixels
defin
where
Ii0 j 0and
is the
rawpeak-probability
value
an
image
pixel,
B
isof&
the
0
0
0sum
0 21
Sraw (d)
the
variance
(d)
are
calculated
as
the
1
if
I
<
SATURATION
mus
ij
i0 raw
j0 = 1 v
i
j
raw
Mi 0 j 0 =
probability:
0
otherwise,
d, i.e..,pixel sensitivity
image, Fplaced
thedistance
normalized
function (flat-field), G is
i0 j 0 is at
x’,y’
i’,j’
is the fractional area
of the
a sub-pixel,
where
quantity mi j i.e.,
is a boolean for a pre-defined bad pixel mask.
0 0
0
0
easily obtain thepline
image
indexes
(i, j) from
sub-pixel
indexes
(i
,
0 ⇥the
1
= ppixel
⇥ pglobalmax
(C
?
),
localmax
0
0
N x Ny
1X
X
31 and 32.
, f i0 j 0 A
ASsraw=(d) = @
Ymeaning
samp
samp
where plocalmaxThe
is output
a boolean
valueis with
same
in equation
spectrum
given the
byX
an
array
of spectralaselements,
where
S of(d)
theflux
variance
calculated
as the sumwithin
of sub-pixels
bility Flux
p
given
value S(y ) are
being
the maximum
a give
Elem Raw
in
e-a and
globalmax,iraw
(yj
2
j
raw (d)
i0
Nx0
Ny0
i0
j0
j0
X
p as:
at A/2),
distance
i.e..,aperture
A/2 < yplaced
< yj +
ford, 2an
A,X
calculated
2
(2 e As + fi0 j 0 ),
raw (d) =
0
Nx0
Ny0
XX
1
2lying within
where fStandard
is given
by
equation
50
and
the
sum
is
calculated
for
sub-pixels
1. Extract
or
Raw
spectrum
(S
)
and
respective
variances
(
i0 j 0 given
std/raw
std/raw ), as
by:
the extraction
aperture.
function FITd represents a fit along dispersion direction
described
in sections
3.1.1 The
and 3.1.2.
1 in section 2.5.
analogously to the fit used for the elements of the 2D0
IP, as described
ˆi0 j 0 for each sub-pixel (i0 , j 0 ) is
2. Construct spatial profile ( ), where the fit profile data B
C
C
B
3.given
Revise
by:variance estimates:
ˆi0 j 0 (d) = FITd B fi0 j 0 C ,
(5
C
B
0 N0
N
Px Py i A
@
Nx0 Ny0 h
0
1
XX
p
fi0 j 0
2
2
ˆ
(53)
(2 e + bi0 j 0 ) As + i0 j 0 S
j 0prev ,
new =
B
C
i0
j0
B fi0 j 0 C
ˆi0 jby
C , is calculated for sub-pixels
0 (d)equation
(52) lying with
= FITd B
where fi0 j 0 is given
50 and
the
sum
0
C
B
0
N
N
Px Py
Arepresents a fit along dispersion directio
@
the
extraction
aperture.
The
function
FIT
0
0
where Sprev is the flux previously calculated using feither
i j d Standard/Raw (first iteration)
j0
the fit used
for thei0elements
of the 2D IP, as described in section 2.5
or Optimalanalogously
(in further to
iterations)
method.
Extracción óptima
Horne et al. (1986) & Marsh et al. (1989)
0 is given by equation 50 and the sum is calculated for sub-pixels lying within
where fi0 j3.
Revise
variance
– 42 –sub-pixel values are identified
4. Mask optimal
outliers.
In estimates:
this step the highly deviant
the extraction aperture. The function FITd represents
a fit along dispersion direction
0
and ignored in the flux calculation (step
in the refinement of the spatial
profile
Nx0 5)Nyand
i
h
X
X
analogously to the fit used for the2 elements of the 2D
IP, asp
described in section 2.5.
2
ˆ
(step 2 - above second iteration).
So, the mask(2function
M 0 0 is updated
as follows:
(5
new =
e + bi0 j 0 ) i jAs + i0 j 0 Sprev ,
i0
j0
3. Revise variance estimates:
Ny0 ⇣
⌘
Nx0 P
P
ˆi0 j 0 fi0 j 0 / 2
(
0
0
N
new
N
y h
–
42
x –X
2
2 (first iteratio
i
ˆiStandard/Raw
X
p
0
0
where
Sifprev
the
flux previously
calculated
using
0 <
0j
0 =
0either
0
0 j 0 Sprev ) < ⌘
1
I2 i0 jis
SATURATION
&
m
1
&
(f
i
j
i
i
j
new
2 S
Mi 0 j 0 =
(53)
=
(2
+ bopt
) ANs 0+N 0ˆi⇣0 j 0 Sprev , ⌘
i0 j 0 =
new
e
or0Optimal
(in
further
iterations)
method.
y
otherwise,0 0
Px P ˆ2
i
j
2
/
0
0
new
ij
(54)
0
0
i
j
4. Mask optimal outliers. In this step the highly deviant sub-pixel values are identifi
2 (first iteration)
Ny0
Ny0 ⇣
⌘
Nx0 units
where
theinput
flux
previously
calculated
using
either
Standard/Raw
Nx0 P
whereSprev
⌘ isisand
an
threshold
which
defines
the
cuto↵
in
of
. For ESPaDOnS
P
P
P
ignored in theˆ flux calculation
(step
5)
and
in
the
refinement
of the spatial profi
2
ˆ
0
0
0 j 0 f i0 j 0 /
ij
i⇡
new means a 2.5 cut.
orwe
Optimal
further
iterations)
method.
use as (in
default
the
value
⌘
7,
which
(step 2 - above
second
iteration).
So,
the
mask
function
Mi0 j 0 is updated as follows:
0
0
0
0
i j
i j
2
Sopt =
(55)
opt = N 0 N 0 ⇣
0 N0 ⇣
⌘
⌘
N
y
y
2
Px P
4.5.Mask
optimal
outliers.
In
this
highly deviant
sub-pixel
values
are identified
Extract
optimal
spectrum
Soptˆstep
variances
, Px P
2
2andthe
opt
2
2
ˆ
i0 j 0 / new
i0 j 0 / new
(
and ignored in the flux calculation
(step 5) and in thei0 refinement
of the spatial profile2
j0
i0 j 0
1
if Ii0 j 0 < SATURATION & mi0 j 0 = 1 & (fi0 j 0 ˆi0 j 0 Sprev ) < ⌘ 2
0
0
new
Extracción de imágenes de calibración
flux (e-/elem)
40
Bias - order 35
Raw Sum
Optimal
Flat - order 35
Raw Sum
Optimal
Th-Ar - order 35
Raw Sum
Optimal
Fabry-Perot - order 35
Raw Sum
Optimal
20
0
-20
flux (e-/elem)
-40
300000
200000
100000
flux (e-/elem)
0
100000
50000
flux (e-/elem)
0
200000
100000
0
0
500
1000
1500
2000
2500
d (pixels)
3000
3500
4000
4500
Extracción en vários niveles de flujo
~SNR per pixel
flux (e-/elem)
flux (e-/elem)
flux (e-/elem)
flux (e-/elem)
200
Raw Sum
Optimal
orders 57 and 58 - faint target
150
orders Faint target: V=13.2, texp=1000 s
min
max
<1
2
1
250
2
1000
50
3000
59
58
100
50
0
orders 32 and 33 - faint target
Raw Sum
Optimal
7500
5000
33
32
2500
0
Raw Sum
Optimal
orders 57 and 58 - bright target
20000
Bright target: V=3.4, texp=10 s
59
58
10000
0
orders 32 and 33 - bright target
Raw Sum
Optimal
75000
50000
33
32
25000
0
0
500
1000
1500
2000
2500
d (pixels)
3000
3500
4000
4500
Extracción de imágenes saturadas
Optimal Extraction
Raw Extraction
flux (e-/elem) + order number * 3e5
1.05e+07
1e+07
9.5e+06
9e+06
0
500
1000
1500
2000
2500
3000
spectral element index
3500
4000
4500
Resolución espectral
Polar mode (POL)
190000
190000
Beam 0
Beam 1
(FP IP) * (FITS Image) POL
(Th-Ar IP) * (FITS Image) POL
Libre-Esprit POL
Na doublet model R=80000
180000
170000
170000
160000
160000
150000
150000
140000
130000
140000
120000
110000
110000
589.7
100000
∆v ≈ -22 km/s
589.6
589.5
589.4
589.3
RV_corr = -13.6 km/s
RV_vega = -13.9 km/s
∆RV = -27.5 km/s
130000
RV_corr = -13.7 km/s
RV_vega = -13.9 km/s
∆RV = -27.6 km/s
120000
100000
Beam 0 - SP2
(FP IP) * (FITS Image) SP2
(Th-Ar IP) * (FITS Image) SP2
Libre-Esprit SP2
Na doublet model R=80000
180000
flux (e-/elem)
flux (e-/elem)
Star-only mode (SP2)
589.2
(nm)
589.1
589
588.9
588.8
589.7
Beam 1
Beam 0
∆v ≈ -26 km/s
589.6
589.5
589.4
589.3
589.2
(nm)
589.1
589
588.9
588.8
Beam 0
Figure 6. Sample of ThAr spectra (up) and flat field images (down) obtained in the two- (left) and four-slice modes (right). Note
how clearly the sliced images can be seen on the ThAr emission lines.
La primera luz de GRACES
Figure 7 shows the raw 2D spectrum of GRACES first light, when the A3 star HIP57258 (V=9.00 mag) was observed on
May 6, 2014. We see that is covers
the optical
band from
~405 nm tode
~1.032014
µm (i.e. from order 58 to order 22), however
en
6
de
Mayo
the spectrum gets very dim bluer than 420 nm. The spectrum is continuous (gap-less) until 922.5 nm, and small 1-2 nm
4.2 Raw spectra
gaps appear between the last 3 orders.
Figure 7. Raw 2D spectrum of the A3 star HIP 57258, also the GRACES first light!
La primera luz de GRACES
6 de Mayo de 2014
35 and 36 (central wavelength at 648 nm and 630 nm, respectively) for both instrument modes as indicated in the
figure.
e. Aperture: this step performs measurements of the tilt angle of a rectangular aperture for extraction. It uses the
instrument profile to measure the tilt angle that maximizes the flux fraction inside the aperture. The calibrated
aperture is aligned with the monochromatic image of the pseudo-slit, allowing unbiased flux measurements of each
spectral element. Bottom panels in Figure 9 show the instrument profiles with the measured extraction apertures
masked out. The average tilt measured for Star-only mode is −2.64±0.07 degrees and for Star+Sky mode is 1.63±0.04 degrees.
Resolución Espectral
Figure 10. Left panels show measurements of the RMS (in red) and median (in green) residual wavelength of matched spectral
lines, where top panel is Star-Only mode and bottom panel is Star+Sky mode. Right panels show measurements of spectral
14
10
Two-slice mode
19mag
20mag
6
21mag
2
500
700
S/N after 1 hour
S/N after 1 hour
Sensibilidad Instrumental
7
5
Four-slice mode
19mag
3
20mag
21mag
1
GRACES
vs HIRES
900
0
500
700
900
Wavelength (nm)
Wavelength
(nm)
Figure 13. S/N reached for a flat spectrum of different magnitudes after a 1h exposure in the two- (left) and the four-slice mode
(right).
Wavelength (nm)
Wavelength
(nm)
Figure 13. S/N reached for a flat spectrum of different magnitudes after a 1h exposure in the two- (left) and the four-slice mode
(right).
Desempeño
Figure 14. Up: Total flux of the Feige66 extracted spectrum in the two- (left) and the
four-slice
modeet(right)
after
a 1h
exposure
Source:
A-N Chene
al. SPIE
2014
paper
OPORTUNIDAD CIENTÍFICA
Espectroscopía de alta resolución
• Abundancia química de alta precisión.
• Atmosferas estelares y planetarias.
• Parámetros estelares: rotación, turbulencia,
temperatura, gravedad de la superficie, etc.
• Actividad magnética en la fotosfera.
• Velocidades radiales - asociaciones, sistemas
binarios, dinámica.
• Líneas de emisión - nebulosas, galaxias, discos,
etc.
•…
GRACES
15 años de ciência con Gemini - La Plata, 02 de Junio de 2015.
EJEMPLO 1: SISTEMA 16 CYG
VELOCIDAD RADIAL Y ABUNDANCIA
Señales de la formación de planetas gigantes
Tucci Maia, M., Meléndez, J. & Ramírez, I.
ApJL, v. 790, i. 2
COCHRAN W., HATZES A., BUTLER P. & MARCY G.
GRACES
Apj., 483, 457
15 años de ciência con Gemini - La Plata, 02 de Junio de 2015.
EJEMPLO 2: AB DORADUS
VELOCIDAD RADIAL ABSOLUTA
BANYAN: Bayesian Analysis for Nearby Young AssociatioNs
Cálculo de la probabilidad de miembros de asociaciones de estrellas (sin información fotometría)
GRACES
Malo et al. (2013)
15 años de ciência con Gemini - La Plata, 02 de Junio de 2015.
EJEMPLO 3: VENUS
Variaciones en la velocidad de los vientos (z ~70km)
PI: Thomas Widemann
Método: desplazamiento Doppler de las líneas solares a la luz dispersada por
partículas de las nubes en movimiento - los datos utilizados para restringir los
modelos de circulación global
Primera evidencia de un flujo meridional en nubes altas en Venus.
Variaciones rápidas de la velocidad del viento reflejan
fuerte dependencia con la latitud y la hora local
EJEMPLO 4: NGC 1624-2 OF?CP TYPE STAR
(Wade, G.A. et al.) Mon.Not.Roy.Astron.Soc. 425 (2012)
SED análisis:
NGC 1624-2 es una estrella de secuencia principal
de masa M=30 Ms, temperatura efectiva de 35000 K
y log g = 4.0 +- 0.2
ULTIMAS NOTICIAS
Instalación de la fase 2 de GRACES acaba de
terminar y el comisionamiento está a punto de
comenzar en junio
MUCHAS GRACIAS!
GRACES
15 años de ciência con Gemini - La Plata, 02 de Junio de 2015.