No Slide Title
Download
Report
Transcript No Slide Title
Spectroscopy principles
Jeremy Allington-Smith
University of Durham
Contents
• Reflection gratings in low order
– Spectral resolution
– Slit width issues
•
•
•
•
•
•
Grisms
Volume Phase Holographic gratings
Immersion
Echelles
Prisms
Predicting efficiency (semi-empirical)
Generic spectrograph layout
f2
Focal ratios
defined as
Fi = fi / Di
Detector
Camera
fT
D2
Grating
Slit
DT
s
Telescope
D1
W
f1
Collimator
Grating equation
n1
n2
A
• Interference condition:
B
path difference between AB and A'B'
A’
B’
• Grating equation:
m n1 sin n2 sin
a
where
1
a
• Dispersion:
d cos
d
m
d d d cos
dx d dx m f 2
f2
d
dx
"Spectral resolution"
• Terminology (sometimes vague!)
– Wavelength resolution d
– Resolving power R d
• Classically, in the diffraction limit,
Resolving power = total number of rulings
x spectral order
*
I.e. R mW
Total grating length
• But in most practical cases for astronomy
( < /DT), the resolving power is determined
by the width of the slit, so R < R*
d
Spectral resolution
• Spectral resolution:
cos
d
d s'
s'
m f 2
dx
• Projected slit width:
f1
Conservation of Etendue (nAW)
s
F2
s
s
s' s s
'
F1
F
sD cos
d d s' cos s 2 1
dx
m f 2
F1
Image of slit
on detector
mD2 f1
Camera focal
length
slit
f2
D1
D2
'
s'
Image
of slit
Resolving power
• Illuminated grating length:
D2
W
cos
• Spectral resolution (width)
s
d
mF1W
• Resolving power:
Collimator
focal ratio
– expressed in laboratory terms
R
mF1W
d
s
– expressed in astronomical terms
R
mW
DT
since
s fT
and
fT
f
1 FT F1
DT D1
Physical
slitwidth
Grating
length
Angular Telescope
slitwidth
size
Size of spectrograph must scale with telescope size
Importance of slit width
• Width of slit determines:
– Resolving power (R) since R = constant
– Throughput (h)
• Hence there is always a tradeoff
between throughput and spectral information
• Function h() depends on Point Spread Function (PSF)
and profile of extended source
– generally h() increases slower than 1 whereas R -1
so hR maximised at small
• Signal/noise also depends on slit width
– throughput ( signal)
– wider slit admits more sky background ( noise)
Signal/noise vs slit width
• For GTC/EMIR in K-band (Balcells et al. 2001)
SNR falls as slit
includes more
sky background
Optimum
slit width
Anamorphism
Output
angle
• Beam size in dispersion direction:
D2 W cos
• Beam size in spatial direction:
D1 W cos
D
cos
A 2
D1 cos
M x D2
A
M D1
• Anamorphic factor:
• Ratio of magnifications:
– if < , A > 1, beam expands
• W increases R increases
• image of slit thinner oversampling
worse
– if > , A < 1, beam squashed
Input
angle
D1
D2
D1
D2
• W reduces R reduces
• image of slit wider oversampling better
– if = , A = 1, beam round
• Littrow configuration
Input
Output
dispersion
Generic spectrograph layout
f2
Fi = fi / Di
Detector
Camera
fT
D2
Grating
Slit
DT
s
Telescope
D1
W
f1
Collimator
Blazing
b = active width
of ruling (b a)
• Diffracted intensity:
Single slit
diffraction
sin 2 N sin 2
I
2
2
sin
/b
Intensity
Interference
pattern
Blaze function
/a
F phase difference between adjacent rulings
= phase difference from centre of one ruling to its edge
-1
0
Spectral order, m
Facet
normal
Shift envelope peak to m=1
• Blaze condition
i
specular reflection off grooves:
2 also -
sin x sin y 2 sin
x y
x- y
cos
2
2
r
m sin sin 2 sin cos
B
since
1
2
Grating
normal
2
Efficiency
Efficiency vs wavelength
• Approximation valid for a >
max(m) = B(m=1)/m
• Rule-of-thumb:
40.5% x peak at
- -
m
B
1
(large m)
– reduction in efficiency with
increasing order
0 .8
2
3
4
6 5
0 .9
• Sum over all orders < 1
Wavelength
m e d iu m re s o lu tio n g ra tin g o rd e rs 2 -6
0 .7
Efficie ncy
m=2
IS A A C g ra tin g e ffie n c y (fro m E S O E T C )
2m B
2m B
and -
2m - 1
2m 1
B
m=1
0 .6
0 .5
0 .4
0 .3
0 .2
0 .1
0
0 .5
1
1 .5
2
W a v e le n g t h ( u m )
(See: Schroeder, Astronomical Optics)
2 .5
3
Order overlaps
Effective passband
in 1st order
Don't forget
higher orders!
Intensity
1st order
blaze profile
m=1
m=2
Passband
in 2nd order
0
(2nd order)
0
L
Zero order
matters
for MOS
2nd order
blaze profile
Passband
in zero
order
m=0
1st order
First and second
orders overlap!
C 2L U
L
2U
U
Wavelength in first
order marking
position on detector
in dispersion direction
(if dispersion ~linear)
Order overlaps
2nd order
Detector
dispersion
Zero
order
1st order
To eliminate overlap between 1st and 2nd order
• Limit wavelength range incident on detector using
passband filter or longpass ("order rejection") filter
acting with long-wavelength cutoff of optics or
detector (e.g. 1100nm for CCD)
• Optimum wavelength range is 1 octave (then 2L = U)
• Zero order may be a problem in multiobject spectroscopy
Predicting efficiency
• Scalar theory approximate
– optical coating has large and unpredictable effects
– grating anomalies not predicted
– Strong polarisation effect at high ruling density
(problem if source polarised or for spectropolarimetry)
• Fabricator's data may only apply to Littrow ( 0)
– convert by multiplying wavelength by cos(/2)
– grating anomalies not predicted
• Coating may affect grating properties in complex
way for large (don't scale just by reflectivity!)
• Two prediction software tools on market
– differential
– integral
CCD mosaic
(6144x4608)
GMOS optical system
Mask field (5.5'x5.5')
Detector (CCD
mosaic)
Science fold
mirror field (7')
Masks and
Integral Field Unit
From
telescope
Example of performance
•
Absolute efficiency of fitted GMOS gratings
(Littrow, unpolarised)
GMOS grating set
– D1 = 100mm, = 50
•
•
– DT = 8m, = 0.5"
90
– m = 1, 13.5mm/px
70
Intended to overcoat
all with silver
80
Efficiency (%)
•
100
R831
60
B600
50
R600
40
R400
30
R150
20
Didn't work for those
with large groove
angle - why?
Actual blaze curves
differed from scalar
theory predictions
B1200
10
0
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
Wavelength (microns)
GMOS
grating
name
B1200
R831
B600
R600
R400
R150
Coating
Al
Al
Ag
Al
Ag
Ag
Ruling
density,
(mm-1)
1200
831
600
600
400
150
Groove
Nominal
angle,
RB
(deg) B(nm)
17.5
463
3744
19.7
757
4396
8.6
461
1688
17.5
926
3744
9.7
764
1918
3.4
717
631
Actual
B(nm)
490
750-900
540
950
940
670-740
Dispersion,
d/dx
(nm/pixel)
0.023
0.034
0.047
0.047
0.085
0.174
Grisms
• Transmission grating
attached to prism
• Allows in-line optical
train:
– simpler to engineer
– quasi-Littrow
configuration - no
variable anamorphism
• Inefficient for
> 600/mm due to
groove shadowing
and other effects
D1
d
nG
nR
n'
Grism equations
• Modified grating equation:
m n sin n' sin
• Undeviated condition:
mU (n - 1) sin
n'= 1, -
= phase difference from
centre of one ruling to its edge
• Blaze condition:
0 B = U
• Resolving power
R
mW
DT
W
D1
(same procedure as for grating)
cos
(n - 1) tan D1
R
DT
Volume Phase Holographic gratings
• So far we have considered surface relief gratings
• An alternative is VPH in which refractive index
varies harmonically throughout the body of the
grating: ng ( x, z) ng ng cos2 g (x sin z cos )
• Don't confuse with 'holographic' gratings (SR)
• Advantages:
–
–
–
–
Higher peak efficiency than SR
Possibility of very large size with high
Blaze condition can be altered (tuned)
Encapsulation in flat glass makes more robust
• Disadvantages
– Tuning of blaze requires bendable spectrograph!
– Issues of wavefront errors and cryogenic use
VPH configurations
• Fringes = planes of
constant n
• Body of grating made
from Dichromated
Gelatine (DCG) which
permanently adopts
fringe pattern generated
holographically
• Fringe orientation allows
operation in transmission
or reflection
/2
2-2
VPH equations
• Modified grating equation:
m sin sin
• Blaze condition:
mB 2ng sin g 2 sin
= Bragg diffraction
• Resolving power:
• Tune blaze condition by
tilting grating ()
• Collimator-camera angle
must also change by 2
mechanical complexity
ng sin g sin
mW m D1
R
DT
DT cos
VPH efficiency
• Kogelnik's analysis when:
2d g2
• Bragg condition when:
• Bragg envelopes (efficiency FWHM):
n g d .
ng
. > 10
2
1
1
n g
– in wavelength:
g tan g
g tan g d
– in angle:
1
gd
• Broad blaze requires
– thin DCG
– large index amplitude
• Superblaze
Barden et al. PASP 112, 809 (2000)
VPH 'grism' = vrism
• Remove bent geometry, allow in-line optical layout
• Use prisms to bend input and output beams while
generating required Bragg condition
n1
ng
n1
d
D1
R
mW m
D1 (1 tand tan )
DT
DT
sin
n1
d - arcsin
Limits to resolving power
• Resolving power can increase as m, and W
increase for a given wavelength, slit and telescope
Grating
parameters
R
D
mW
1
DT
DT
sin sin
cos
Geometrical
factors
• Limit depends on geometrical factors only increasing or m will not help!
• In practice, the limit is when the output beam
overfills the camera:
– W is actually the length of the intersection between
beam and grating plane - not the actual grating length
– R will increase even if grating overfilled until diffractionlimited regime is entered ( > DT)
Limits with normal gratings
• For GMOS with
= 0.5", DT = 8m,
D1 =100mm,
=50
• R and plotted as
function of
•
A(max) = 1.5 since
D2(max) = 150mm
R(max) ~ 5000
Grating
A
B
C
D
H
IA
X1
X2
X3
(mm )
(deg)
1200
831
600
600
400
158
2400
1800
1200
17.5
19.7
8.6
17.5
9.7
3.6
21.0
26.7
17.5
-1
(deg)
0.0
0.0
0.0
0.0
0.0
0.0
30.0
30.0
30.0
B
463
757
461
926
764
720
444
724
724
RB
3744
4396
1688
3744
1918
668
7043
9572
5372
Normal SR
gratings
Simultaneous
range
Immersed gratings
• Beat the limit using a prism to squash the output
beam before it enters the camera:
D2 kept small while W can be large
• Prism is immersed to prism using an optical
couplant (similar n to prism and high transmission)
• For GMOS R(max) ~ doubled!
camera
• Potential drawbacks:
– loss of efficiency
– ghost images
– but Lee & Allington-Smith
(MNRAS, 312, 57, 2000)
show this is not the case
prism
collimator
grating
Immersed 1800/mm grating
with 35 deg prism, 630-682nm
Limits with immersed gratings
• For GMOS with
= 0.5", DT = 8m,
D1 = 100mm
• R and plotted as
function of
• With immersion
R ~ 10000 okay
with wide slit
Grating
A
B
C
D
H
IA
X1
X2
X3
(mm-1)
(deg)
1200
831
600
600
400
158
2400
1800
1200
17.5
19.7
8.6
17.5
9.7
3.6
21.0
26.7
17.5
(deg)
0.0
0.0
0.0
0.0
0.0
0.0
30.0
30.0
30.0
B
463
757
461
926
764
720
444
724
724
RB
3744
4396
1688
3744
1918
668
7043
9572
5372
Immersed
gratings
Echelle gratings
• Obtain very high R (> 105) using very long grating
• In Littrow:
R
mW 2 D1
tan
DT
DT
Groove
angle
• Maximising requires large m since m = 2sin
• Instead of increasing , increase m
• Echelle is a coarse
grating with large
groove angle
• R parameter = tan
(e.g R2 = 63.5)
D1
W
Multiple orders
– use narrow passband
filter to isolate one
order at a time
– cross-disperse to fill
detector with many
orders at once
10
555
500
11
Cross-dispersion
• Many orders to
cover desired :
Free spectral range
/m
• Orders lie on top of
each other:
(m) = (n) (n/m)
• Solution:
555
454
12
417
13
385
14
357
15
333
16
333
312
312
300
Primary dispersion
Cross dispersion may use prisms
or low dispersion grating
Echellette example - ESI
Sheinis et al. PASP 114, 851 (2002)
Prisms
• Useful where only low resolving power is required
• Advantages:
– simple - no rulings! (but glass must be of high quality)
– multiple-order overlap not a problem - only one order!
• Disadvantages:
– high resolving power not possible
– resolving power/resolution can vary strongly with
L
D1
D2
t
Prism, index n
Dispersion for prisms
• Fermat's principle:
• Dispersion:
tn 2L cos
d d dn
d dn d
d
d
-2
dn
dn
- - 2
dn - 2 L sin - 2 D1
d
t
t
dn
1 dn D1
d
2 d
t
L
D1
D2
t
Prism, index n
d D1 d
d
t dn
Resolving power for prisms
d
d d
d
• Basic definitions:
• Conservation of Etendue:
Angular
slitwidth
DT d D2
Telescope
aperture
• Result:
Angular width
of resolution
element
on detector
Beam
size
R
• Comparison of grating and prism:
Disperser
'length'
d
Angular
dispersion
D d DT DT d
d 1
t dn
t dn D2
R *prism t
R
dn
d
*
Rgrat
mW
'Ruling
density'
dn
t
DT d DT
*
R prism
Prism example
A design for Near-infrared spectrograph* of NGST
• DT = 8m, = 0.1", D1 = D2 = 86mm, 1 < < 5mm
• R 100 required
Raw refractive index
Slit plane
data for sapphire
Collimator
2.3
0
-0.02
2.29
n
-0.06
n (crystan)
2.28
-0.08
dn/dlambda
(Crystan)
2.27
-0.1
-0.12
Detector
2.26
Camera
dn/dlambda
Double-pass
prism+mirror
-0.04
-0.14
-0.16
2.25
-0.18
2.24
-0.2
1
2
3
4
5
Wavelength (um)
* ESO/LAM/Durham/Astrium et al. for ESA
Prism example
(contd)
• Required prism
thickness,t:
– sapphire: 20mm
– ZnS/ZnSe: 15mm
• Uniformity in d
or R required?
• For ZnS:
2 D1
nt
n 2.26 = 75.3
t
t
sin
sin
2
2 L 2 D1
= 12.9
1000
arctan
900
R
L D1 sin
Resolving power
800
R for t=20mm ZnSe
700
R for t=20mm ZnS
600
R for t=15mm Sappphire
500
Fit for ZnS
400
Fit for sapphire
300
200
100
0
1
2
3
Wavelength (m m)
4
5
Appendix: Semi-empirical
efficiency prediction for
classical gratings
Efficiency - semi-empirical
• Efficiency as a function of depends mostly on
• Different behaviour depends on polarisation:
P - parallel to grooves (TE)
S - perpendicular to grooves (TM)
• Overall peak at = 2sin (for Littrow examples)
• Anomalies (passoff) when light diffracted from an
order at = /2 light redistributed into other orders
2
– discontinuities at a
2m - 1
• Littrow: symmetry m 1-m
(Littrow only)
m
a
-3
0.29
-2
0.40
-1
0.67
0
-
+1
-
+2
0.67
• Otherwise: no symmetry ( depends on m,) double anomalies
– Also resonance anomalies - harder to predict
+3
0.40
Efficiency - semi-empirical
(contd)
Different regimes for blazed (triangular) grooves
< 5
obeys scalar theory, little polarisation effect (P S)
5 < < 10
S anomaly at 2/3 , P peaks at lower than S
10 < < 18
various S anomalies
18 < < 22
anomalies suppressed, S >> P at large
22 < < 38
strong S anomaly at P peak, S constant at large
> 38
S and P peaks very different, efficient in Littrow only
NOTE
Results apply to
Littrow only
From: Diffraction
Grating Handbook,
C. Palmer, Thermo RGL,
(www.gratinglab.com)