Title: Spectroscopy principles
1Spectroscopy principles
- Jeremy Allington-Smith
- University of Durham
2Contents
- Reflection gratings in low order
- Spectral resolution
- Slit width issues
- Grisms
- Volume Phase Holographic gratings
- Immersion
- Echelles
- Prisms
- Predicting efficiency (semi-empirical)
3Generic spectrograph layout
Focal ratios defined as Fi fi / Di
4Grating equation
- Interference condition
- ? path difference between AB and A'B'
- Grating equation
- Dispersion
f2
dx
db
5"Spectral resolution"
- Terminology (sometimes vague!)
- Wavelength resolution dl
- Resolving power
- Classically, in the diffraction limit,
- Resolving power total number of rulings x
spectral order - I.e.
- But in most practical cases for astronomy (c lt
l/DT), the resolving power is determined by the
width of the slit, so R lt R
Total grating length
6Spectral resolution
- Spectral resolution
- Projected slit width
- Conservation of Etendue (nAW)
-
- ? ?
-
- ?
Image of slit on detector
Camera focal length
7Resolving power
- Illuminated grating length
- Spectral resolution (width)
- Resolving power
- expressed in laboratory terms
- expressed in astronomical terms
- since and
Collimator focal ratio
Physical slitwidth
Grating length
Angular slitwidth
Telescope size
8Importance of slit width
- Width of slit determines
- Resolving power (R) since Rc constant
- Throughput (h)
- Hence there is always a tradeoff
- between throughput and spectral information
- Function h(c) depends on Point Spread Function
(PSF) and profile of extended source - generally h(c) increases slower than c1 whereas
R ? c-1 so hR maximised at small c - Signal/noise also depends on slit width
- throughput (? signal)
- wider slit admits more sky background (? noise)
9Signal/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
10Anamorphism
Output angle
- Beam size in dispersion direction
- Beam size in spatial direction
- Anamorphic factor
- Ratio of magnifications
- if b lt a, A gt 1, beam expands
- W increases ? R increases
- image of slit thinner ? oversampling worse
- if b gt a, A lt 1, beam squashed
- W reduces ? R reduces
- image of slit wider ? oversampling better
- if b a, A 1, beam round
- Littrow configuration
Input angle
11Generic spectrograph layout
Fi fi / Di
12Blazing
b active width of ruling (b ? a)
- Diffracted intensity
- Shift envelope peak to m1
- Blaze condition
- specular reflection off grooves
- also
- ?
- since
Single slit diffraction
Interference pattern
F phase difference between adjacent rulings q
phase difference from centre of one ruling to its
edge
13Efficiency vs wavelength
- Approximation valid for a gt l
- lmax(m) lB(m1)/m
- Rule-of-thumb
- 40.5 x peak at
- ? (large m)
- Sum over all orders lt 1
- reduction in efficiency with increasing order
2
3
4
5
6
(See Schroeder, Astronomical Optics)
14Order overlaps
Effective passband in 1st order
Don't forget higher orders!
Intensity
1st order blaze profile
m1
First and second orders overlap!
m2
Passband in 2nd order
Zero order matters for MOS
2nd order blaze profile
Passband in zero order
m0
Wavelength in first order marking position on
detector in dispersion direction (if dispersion
linear)
1st order
0
lU
2lL
2lU
lL
lC
(2nd order)
0
lL
lU
15Order overlaps
Detector
1st order
2nd order
Zero 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 2lL
lU) - Zero order may be a problem in multiobject
spectroscopy
16Predicting 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 (Y
0) - convert by multiplying wavelength by cos(Y/2)
- grating anomalies not predicted
- Coating may affect grating properties in complex
way for large g (don't scale just by
reflectivity!) - Two prediction software tools on market
- differential
- integral
17GMOS optical system
18Example of performance
- GMOS grating set
- D1 100mm, Y 50?
- DT 8m, c 0.5"
- m 1, 13.5mm/px
- Intended to overcoat all with silver
- Didn't work for those with large groove angle -
why? - Actual blaze curves differed from scalar theory
predictions
19Grisms
- Transmission grating attached to prism
- Allows in-line optical train
- simpler to engineer
- quasi-Littrow configuration - no variable
anamorphism - Inefficient for r gt 600/mm due to groove
shadowing and other effects
20Grism equations
- Modified grating equation
- Undeviated condition
- n' 1, b -a f
- Blaze condition q 0 ? lB lU
- Resolving power
- (same procedure as for grating)
q phase difference from centre of one ruling to
its edge
21Volume 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 - Don't confuse with 'holographic' gratings (SR)
- Advantages
- Higher peak efficiency than SR
- Possibility of very large size with high r
- 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
22VPH 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
23VPH equations
- Modified grating equation
- Blaze condition
- Bragg diffraction
- Resolving power
- Tune blaze condition by tilting grating (a)
- Collimator-camera angle must also change by 2a
? mechanical complexity
24VPH efficiency
- Kogelnik's analysis when
- Bragg condition when
- Bragg envelopes (efficiency FWHM)
- in wavelength
- in angle
- Broad blaze requires
- thin DCG
- large index amplitude
- Superblaze
25VPH 'grism' vrism
- Remove bent geometry, allow in-line optical
layout - Use prisms to bend input and output beams while
generating required Bragg condition
26Limits to resolving power
- Resolving power can increase as m, r and W
increase for a given wavelength, slit and
telescope - Limit depends on geometrical factors only -
increasing r 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
diffraction-limited regime is entered (l gt cDT)
Geometrical factors
Grating parameters
27Limits with normal gratings
- For GMOS with c 0.5", DT 8m, D1 100mm, Y
50? - R and l plotted as function of a
- A(max) 1.5 since
- D2(max) 150mm ? R(max) 5000
Normal SR gratings
Simultaneous l range
28Immersed 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!
- Potential drawbacks
- loss of efficiency
- ghost images
- but Lee Allington-Smith (MNRAS, 312, 57, 2000)
show this is not the case
29Limits with immersed gratings
- For GMOS with c 0.5", DT 8m, D1 100mm
- R and l plotted as function of a
- With immersion R 10000 okay with wide slit
Immersed gratings
30Echelle gratings
- Obtain very high R (gt 105) using very long
grating - In Littrow
- Maximising g requires large mr since mrl 2sing
- Instead of increasing r, increase m
- Echelle is a coarse
- grating with large
- groove angle
- R parameter tang
- (e.g R2 ? g 63.5)
Groove angle
31Multiple orders
- Many orders to cover desired ll Free spectral
range - Dl l/m
- Orders lie on top of each other
- l(m) l(n)? (n/m)
- Solution
- use narrow passband filter to isolate one order
at a time - cross-disperse to fill detector with many orders
at once
Cross dispersion may use prisms or low dispersion
grating
32Echellette example - ESI
Sheinis et al. PASP 114, 851 (2002)
33Prisms
- 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
34Dispersion for prisms
- Fermat's principle
- Dispersion
35Resolving power for prisms
Angular width of resolution element on detector
- Basic definitions
- Conservation of Etendue
- Result
- Comparison of grating and prism
Angular dispersion
Angular slitwidth
Beam size
Telescope aperture
Disperser 'length'
'Ruling density'
36Prism example
- A design for Near-infrared spectrograph of NGST
- DT 8m, c 0.1", D1 D2 86mm, 1 lt l lt 5mm
- R ? 100 required
Raw refractive index data for sapphire
Collimator
Slit plane
Double-pass prismmirror
Detector
Camera
ESO/LAM/Durham/Astrium et al. for ESA
37Prism example (contd)
- Required prism thickness,t
- sapphire 20mm
- ZnS/ZnSe 15mm
- Uniformity in dl or R required?
- For ZnS
- n ? 2.26 ? a 75.3?
- f 12.9?
38Appendix Semi-empirical efficiency prediction
for classical gratings
39Efficiency - semi-empirical
- Efficiency as a function of rl depends mostly on
g - Different behaviour depends on polarisation
- P - parallel to grooves (TE)
- S - perpendicular to grooves (TM)
- Overall peak at rl 2sing (for Littrow
examples) - Anomalies (passoff) when light diffracted from an
order at b p/2 ? light redistributed into other
orders - discontinuities at (Littrow only)
- Littrow symmetry m? 1-m
- Otherwise no symmetry (rl depends on m,Y) ?
double anomalies - Also resonance anomalies - harder to predict
40Efficiency - semi-empirical (contd)
- Different regimes for blazed (triangular) grooves
- g lt 5? obeys scalar theory, little
polarisation effect (P ? S) - 5 lt g lt 10? S anomaly at rl ? 2/3 , P peaks at
lowerrl than S - 10 lt g lt 18? various S anomalies
- 18 lt g lt 22? anomalies suppressed, S gtgt P at
large rl - 22 lt g lt 38? strong S anomaly at P peak, S
constant at large rl - g gt 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)
rl
ab