MACAO

MAD

Adaptive Optics 101

Lecture 2
CRIRES NACO SINFONI

Overview
1. 2. 3. 4.

5.

Reference material Image formation A statistical description of turbulence Zernike polynomials and Karhunen-Loeve functions Temporal evolution of Zernike co-efficients

Reference Material
Clair Max lectures at [Link] CTIO tutorial at [Link] (A lot of the material in this talk comes from these lectures) Roddiers book - Adaptive optics in astronomy Tysons books - Introduction to Adaptive Optics, Principles of adaptive optics

Image formation (no telescope)

Pupil plane

Source plane

Image plane P

Point Source at P0

r "#

s"#

The disturbance at point P from a point source at point P0 propagating through an opening in an opaque screen is given by the Fresnel-Kirchoff diffraction formula. For astronomical sources, the distance r is significantly larger than the aperture size. If s is also significantly larger than the aperture, one can use the Fraunhofer (far field) approximation to the Fresnel-Kirchoff diffraction formula. With this approximation, the field at point P is the Fourier transform of the pupil transmission function.

Image formation (add the telescope)

Pupil plane

Source plane

Image plane P

r "#

The effect of the lens is to move the diffraction pattern from infinity (Fraunhofer, far-field case) to the focus of the lens. The field at point at point P, which is now a distance f from the pupil plane, is the Fourier transform of the pupil transmission function.

U( u,v )= P =

$ P(x, y)e

"2 # i(xu+yv )

dxdy

Where x and y are distances in the pupil in terms of wavelengths, and u and v are angles that are measured in radians. U and P are complex quantities.

The Airy Function

For a circular disk (e.g., a telescope mirror), the intensity, I(w ) - w2=u2+v2, is
2

$1, 0 < " < a ' P( ",# ) = % ( &0, otherwise)

2 = I(0) [2J1 (2" aw) /2" aw ] 2 I(w) = U(w) = P

!
This is known as the Airy function.

!
The telescope pupil is not a complete disk. There is hole in the middle. So the diffraction pattern of the telescope is not quite the Airy function, but the difference is small.

1.22" /D

" /D

The Imaging Equation and the OTF

The image is a convolution between the brightness distribution of the astronomical object and the PSF.

r I( w ) =

r r r r # O(w")P(w" $ w)dw"

In Fourier space, the convolution becomes a multiplication.

!Where ~ signifies a FT

r r r I ( f ) = O( f ) " P ( f )

r f

is a spatial frequency with units of inverse radians or cycles per radian

r ! P( f )

is called the Optical Transfer Function (OTF)

The Optical Transfer Function (OTF)

The OTF is the Fourier transform of the point spread function (PSF). It is also the autocorrelation function of the pupil. The modulus of the OTF is called the modular transfer function (MTF).

Note how the MTF goes to zero at D/ (~18 arcsec-1 for the VLT), where D is the telescope diameter. This means that higher spatial frequencies (the fine details) are filtered by the system. In the presence of turbulence, the MTF is reduced.

The autocorrelation and coherence functions

The OTF is the autocorrelation function of the amplitude of the electromagnetic wave in the pupil.

%
!

r r r * r "( x #)" ( x # $ x )dx #

In the case of a plane wave from a point source entering a pupil that is aberration free, the OTF is a real function. In the case of statistical fluctuations of the electromagnetic wave due to atmospheric turbulence, the autocorrelation function can be expressed as an ensemble average over all possible realisations. This is called the coherence function

r r r * r "( x ) = #( x $)# ( x $ % x )
coherence function (or OTF)

Turbulence

PSF

Turbulence
Energy is injected into a an eddy of some size - L0 This breaks up into smaller eddies, which then break up inter smaller and smaller energies until Re becomes small. This occurs at lo (the inner scale length or microscale). It is where energy dissipation due to viscosity becomes important. l0 is called the inner scale length and is of the order of a few mm to a few cm L0 is called the outer scale length of the order of a few 10s of m to few 100s of m l0< | r | < L0 is often called the inertial subrange

The structure function

Because of the random nature of turbulence one needs a statistical approach to describe it. Under some assumptions (if the turbulence is homogeneous, isotropic and incompressible), then the mean of the square of the refractive index difference can be written

r r r 2 Dn (r) = [ n( r + r1 ) " n( r1 )] r n( r ) !
is the refractive index

This is called a structure function and it is closely related to the covariance function. The angular brackets signify an ensemble average.

Kolmogorov Turbulence
For Kolmogorov turbulence, the structure function takes on a particular form
2 Dn (r) = Cn r 2 / 3 ; l0 < r < L0

This is known as Obukhovs law Where Cn2 is the refractive index structure constant. It is a measure of the strength ! the turbulence and it is by no means constant. It varies with the of seasons, nightly and hourly and it varies with altitude. The 1/3rd power is seen in nearly all formulae involving active optics It is strictly valid over the inertial sub-range, i.e., l0< | r | < L0 A more general model is the van Karman model

The spatial power spectrum

For Kolmogorov turbulence, The modified von Karman spectrum is

r r $11/ 3 2 "(# ) = 0.033Cn (z) #

r 0.033C "(# ) = r 2 ! (# + # )

2 n 2 11/ 6 0

r % # 2( exp'$ 2 *; # 0 = 2+ /L0 ' # * & m)

Be careful, several definitions for L0 exist.

Also is given several names - wavenumber, spatial frequency, etc.

VLTI

ELT

UTs

3.6m

Tip-tilt jitter
For Kolmogorov turbulence,
2 " #$ = 0.36 %2 D&1/ 3 r0 &5 / 3

No wavelength dependence. Depends weekly on diameter. ! If the jitter for the 8m is 0.5, then the jitter for the laser on the Na layer is ~0.8 (the Launch telescope has a diameter of 0.5m). Combine this with the jitter of the 8m, one gets a jitter of ~1, twice the seeing. Actually, in a sub-pupil of the NACO 14x14 lenslet array, the jitter would be about 1.1.

What happens when the outer scale length is finite?

The coherence function of a turbulent atmosphere

After propagating through a turbulent atmosphere, the coherence function is,

r 1 # D$ ( x ) r "( x ) = e 2 =e

% x (5/3 #3.44 ' * & r0 )

D" (x)
!

is the phase structure function and r0

2" 2 r0 = [0.423( ) X $ C 2 N (h)dh](%3 / 5) #

is the Fried Parameter. x is now a physical displacement

!
!

The phase structure function

For Kolmogorov turbulence rad^2, Units are rad^2, x is a distance

# x &5 / 3 r D" ( x ) = 6.88% ( $ r0 '

Putting it all together

We can separate the OTF into two components. That of the telescope and that of the atmosphere. In Fourier space, this is a multiplication. In normal space this is a convolution.
% x (5/3 $3.44 ' * & r0 )

P = PTelescope " PAtmosphere # PAtmosphere = e

Recall that the Fourier transform of the OTF is the PSF. The PSF is close to a Gaussian (not quite) with FWHM.

# " = 0.98 r0

Phases and wave-fronts

Telescope aperture To convert from a physical displacement of the wavefront, W, to a displacement in the phase, one uses

"

!
Wavefront - a surfaces of constant phase

r 2# r "( x) = W ( x) $
A displacement of 200 nm corresponds to 0.58 radians of phase for light in the K-band. In the V-band, this is 2.28 radians

An example - tilt
Telescope aperture

U(u,v) =

P(x, y)e i" e#2 $ i(ux +vy )dxdy %

Added a phase term to the pupil function

For simple tilt,

" = 2# ax

We can rewrite the above equation as

U(u,v) = !

P(x, y)e"2 # i((u"a )x +vy )dxdy $

A simple tilt at the pupil plane corresponds to a shift in the image plane

Representing the wave front

Power series (Seidel aberrations)
%

"( #,$ ) =
Zernike series

&S
n,m= 0

# n cosm $ + Snm 2 # n sin m $ nm1

"( #,$ ) = & a j Z j ( #,$ )

j=1

Over a circular aperture, the Zernike series are an orthonromal basis. Karhunen-Loeve functions

The first eight Zernike terms - I

Piston Tip and tilt Defocus Astigmatism

Z1 = 1

Z 2 = 2 " cos # ; Z 3 = 2 " sin # Z 4 = 3(2 " 2 #1)

Z 5 = 6 " 2 cos2# ; Z 6 = 6 " 2 sin2#
Z 8 = 8(3" 3 # 2 ")sin $ Z 7 = 8(3" 3 # 2 ")cos $ ;

! Coma

Higher order terms get names as well - trefoil, spherical, ashtray (or tretrafoil), secondary astigmatism, etc. ! Different names, orderings and normalisations are used ! Zernike terms have a radial degree and an azimuthal frequency. E.g. coma has a radial degree of 3 and an a azimuthal frequency of 1

The first twenty-eight Zernikes

Courtesy of Patrick Y. Maeda

 and the effect in the image plane

Courtesy of Patrick Y. Maeda

Useful properties of Zernikes r r r r Over a cicular aperture, the Zernike # W ( x )Z ( x )Z ( x )dx = "
j k

jk

series are an orthonormal basis.

r &1/ $ (x % 1)) W ( x) = ' * (x > 1) + (0 r r "( x) = # a j Z j ( x)

j

Any phase distortion over a circular aperture can be expanded as a sum of Zernike modes.

For Kolmogorov Turbulence, the covariance of the coefficients is cjk are elements of the Noll Matrix The phase variance is

a j ak = c jk (D /r0 ) 5 / 3
# #

" 2 = $ a 2 = $ c jj (D /r0 ) 5 / 3 j
j= 2 j= 2

The Noll Matrix

cjk are elements of the Noll Matrix The matrix is not quite diagonal.

a j ak = c jk (D /r0 ) 5 / 3

AO correction
What happens when we correct the lower order Zernike modes with an AO system perfectly. If we correct the first N modes the phase variance becomes
# 2 "J =

$c
j= J +1

(D /r0 ) 5 / 3 = % J (D /r0 ) 5 / 3 jj
$ 3 /2

J
1 2

"J
1.030 0.582 0.134 0.111 0.0880 0.0648 Focus and astigmatism Tip and tilt

" J # 0.2944J

J > 10

Recall the Marechal approximation

!3
4 5 6

S=e

"#

$2 , #2 < 4

For 0.5 seeing in the K-band, one would get 44% Strehl correcting just the first six Zernikes on a 4m telescope. On an 8m, this would be 7%.

Question: How many modes would you need to correct if you wanted 44% Strehl on an 8m?

Partially corrected images - I

How does the phase structure function and PSF change when we start correcting the lower order Zernike modes with an AO system?

r D" ( x ) # 2$ 2

r "( x ) = e
!

r 1 # D$ ( x ) 2

r $% 2 "( x ) # e

Partially corrected images - II

The partially corrected PSF will consist of a diffraction limited core and a seeing limited halo.

r r r P( u) = (1" S) # Phalo ( u) + S # Pdiff ( u)

!
Intensity

Karhunen-Loeve functions
The Noll matrix is not diagonal. Is there another basis in which the covariance matrix is diagonal? Yes! Karhunen-Loeve functions - linear combination of Zernike modes with the same Azimuthal frequency Why bother? The residual variance as a function of the number of modes corrected decreases more quickly if KL functions are used.

Temporal evolution of Zernike coefficients

The temporal covariance is

Ci (" ) = ai (t)ai (t + " )

The spatial covariance is

! r Where "( k ) is the spatial power spectrum (Weiner-Kinchin theorem)

The Taylor hypothesis of frozen turbulence - the lifetime of a ! turbulent cell is larger than the time in takes to cross the pupil. The ! position of a turbulent cell at time is simply v.

r r r r r Bi ( ") = ai ( r )ai ( r + " ) = #( k )

r r Bi ( ") = Ci (v # )

So the temporal frequency spectrum is related to the spatial power spectrum.

Temporal power spectra - I

r v /D " 10 Hz

r f t " 0.3(n + 1)v /D

n

The band width of the correction should be about the same as the transition frequency, which increases with radial degree - n

Temporal power spectra - II

However, the correction for a single mode is never perfect. The residual variance from, e.g., tip-tilt, may be larger than the uncorrected variance of the higher order mode. This means that one typically wants to have the same bandwidth for all modes. For NACO, we tune the bandwidth in real time. It is called modal optimisation, but we will leave this for a future lecture. And finally, the bandwidth is not the frame rate. Typically, the frame rate = 10 x bandwidth.

Future Lectures
Atmospheric turbulence, image formation, Zernike polynomials, Karhunen-Loeve expansion AO systems, wave-front sensors, calibrating AO systems - interaction matrices, reference slopes, static aberrations, pupil alignment, etc. AO control techniques, transfer functions, loop bandwidth, fitting errors Performance estimation, aliasing, AO simulation tools Methods for measuring 0 , 0 , etc. DIMM, SCIDAR, SODAR, MASS PSF reconstruction We plan to give one lecture every two months over the coming year

Team AO Projects
Voluntary Last one year Challenging If you show commitment, were committed to help, and you will learn a lot about AO
Estimating the real time performance of the MACAOs - (E. Mason) PSF estimation for the ETCs of CRIRES and SINFONI - (E. Valenti) PSF reconstruction from NAOS RTC data Re-evaluate the accuracy of the NACO preparation software Examine the long term performance of NACO Vibration measurement campaigns on NACO, SINFONI and CRIRES Measuring and correcting for static aberrations on NACO (and SINFONI?) Applying static aberrations to CRIRES MACAO to improve the accuracy of radial velocity measurements. Your own idea?

See the Team AO web pages for details