Optical Coatings in LIGO

Physics 208, Electro-optics

Peter Beyersdorf

Document info 1
Optical Coatings in LIGO
Stack of alternating λ/4 layers of high index
Ta2O5 and low index SiO2

Materials chosen for low loss and high

dielectric contrast (nh-nl)

λ/4 thickness maximizes reflectivity for a given

number of layers
Mirror Requirements for LIGO

Mirror reflectivity determines finesse of arm

cavities and should be kept as high as possible
to maximize interferometer response

Brownian thermal noise of the mirror coating

limits noise floor from 30 Hz-300 Hz

Conventional mirror coating design optimizes

reflectivity for a given number of layers, but
what is needed is optimal thermal noise for a
given reflectivity
 Significance of Coating Design

Thermal noise from optical coatings in gravitational

wave detectors

Gregory M. Harry, Helena Armandula, Eric Black, D. R. M. Crooks, Gianpietro Cagnoli,

Jim Hough, Peter Murray, Stuart Reid, Sheila Rowan, Peter Sneddon, Martin M. Fejer,
Roger Route, and Steven D. Penn

Gravitational waves are a prediction of Einstein’s general theory of relativity. These waves are created
by massive objects, like neutron stars or black holes, oscillating at speeds appreciable to the speed of light.
The detectable effect on the Earth of these waves is extremely small, however, creating strains of the
order of 10!21. There are a number of basic physics experiments around the world designed to detect these
waves by using interferometers with very long arms, up to 4 km in length. The next-generation inter-
ferometers are currently being designed, and the thermal noise in the mirrors will set the sensitivity over
much of the usable bandwidth. Thermal noise arising from mechanical loss in the optical coatings put on
the mirrors will be a significant source of noise. Achieving higher sensitivity through lower mechanical
loss coatings, while preserving the crucial optical and thermal properties, is an area of active research
right now. © 2006 Optical Society of America
OCIS codes: 310.1620, 310.6870, 350.1270.

1. Introduction
Isaac Newton’s description of gravity was improved Gmv2!"rc4# $ 10!21 % &% & %
m
M!
v
c
2 100 Mpc
r
,& (1)
upon in 1915 by Albert Einstein when the latter’s
general theory of relativity was published. This the- where G is Newton’s gravitational constant, m is the
ory allows for oscillations in space–time, caused by mass of the source, v is the velocity of the source, r is
motions of masses analogous to electromagnetic the distance from the detector to the source, c is the

“In the most sensitive bandwidth, between 40 Hz and a

waves arising from moving charges in Maxwell’s the-
for the 4 km long LIGO interferometer in Hanford (left). The few
solidhundred
ory. These oscillations, known as gravitational
waves, create a strain in space–time, so the travel
time for a light beam between two inertial masses
se from September 2002 to August 2004. The design sensitivity
will change as the wave goes by. The size of this
strain is set by the ratio
source. This is the thermal
speed of light, M! is a solar mass, and Mpc is a
megaparsec. To even approach measurable strains,
curve hertz,
is the thermal
design noise; the curves above that show th
astronomical-sized masses moving at appreciable
noise is the dominant noise
fractions of the speed of light are necessary. A typical
gravitational wave at Earth from a source at inter-
of the proposed Advanced LIGO interferometer
mirror faces is shown o
galactic distances is expected to have a strain near
motion of the
10!21 or less. This is roughly a change in length equal
to the width of a human hair over the distance be-

e the difference in the y-axis offset between the two graphs; the Advanced
themselves
G. M. Harry (gharry@ligo.mit.edu), H. Armandula, and E. Blackand LIGO will have
comes primarily lower
from the noise (about
mechanical lossa factor of 1
tween the Sun and the nearest star.
There are a number of experiments1–3 that use in-
terferometry to attempt to detect these waves. A typ-
are with the Laser Interferometer Gravitational-Wave Observa-

al LIGO in most bands. in the optical coatings. The thermal-noise curve in Fig. 3
tory (LIGO) Laboratory; G. M. Harry, with the Massachusetts
Institute of Technology, NW17-161, Cambridge, Massachusetts
01239; H. Armandula and E. Black, with the California Institute of
Technology, 1200 East California Boulevard, Pasadena, California
91125. D. R. M. Crooks, G. Cagnoli, J. Hough, P. Murray, S. Reid,
S. Rowan, and P. Sneddon are with the Department of Physics and
Astronomy, The University of Glasgow, Glasgow G12 8QQ, United
Kingdom. M. M. Fejer and R. Route are with the Edward L. Ginz-
ton Laboratory, Stanford University, Stanford, California
94305. S. D. Penn is with the Department of Physics, Hobart and
f the sensitive bandwidth. A higher-power
William Smith Colleges, Geneva, New York 14456.
Received 1 March 2005; accepted 7 July 2005.
0003-6935/06/071569-06$15.00/0
© 2006 Optical Society of America
educe shot noise at high frequencies. The
time with the mirrors will also be made
ncreasing the finesse of the Fabry–Perot
the most sensitive bandwidth, between
ical Michelson interferometer design with two
perpendicular arms is shown in Fig. 1. The tensor
field of the gravitational wave is most easily detected
by using two perpendicular arms. This is in contrast
to the vector electromagnetic field, in which a single
assumes the same ion-beam-deposited silica-tantala
linear antenna suffices. The mirrors and other optics
of the interferometer hang as pendulums. This gives
the best approximation of a freely falling mass;
coating as was used for the initial LIGO. This noise will set
nearly free in the sensitive direction of the inter-
the ultimate sensitivity, and thus the astronomical
angles of the ofsilica substrate LIGO (no subscript) and th
ferometer but supported against the static gravita-
tional field of the Earth. To increase the signal, many
experiments make each long arm a Fabry–Perot cav-
ity to increase the interaction time with the mirrors. effectiveness, the Advanced interferometers.
1 March 2006 ! Vol. 45, No. 7 ! APPLIED OPTICS 1569
coating forthisstresses
Reducing noise fromperpendicular
the level shown would#!$ and
have big parallel #
payoffs for gravitational wave detection and astronomy.”
to the optic face, respectively. If all the Poisson ratio
are small, Eq. (2) can be approximated as
 Conventional
dielectric mirror coatings
HR mirror coatings have λ/2 periodicity

normally λ/4, high and low index layers

Alternating 3λ/8, λ/8 layers or (n-1)λ/n, λ/n

layer geometry is also possible

First and last layers are often different to

match other requirements

sn.
How to design non periodic coatings

Genetic Algorithm [J.H. Holland (1975)

successfully applied to many constrained
design problems]

Example of one (conservative) optimization:

Minimizing transmittance (<20 ppm);

Minimizing Ta2O5 thickness (<2 μm).

Alternatives: Regular non-periodic coatings

(e.g. pre-fractal) ?
sn.
 Genetic Synthesis

C={x1, x2, x3, … xN}

x1 x2 …
Consider the thickness of each layer in an alternating stack of high
and low index material to be one “gene” in a chromosone describing
the mirror
Create a population of “organisms” each with one chromosone that
has randomized values for each gene
Evaluate fitness of each “organism” as a function of the mirror
reflectivity and tantala thickness for the mirror it represents
Preferentially select most fit organisms for breeding
Mutate random genes
Split chromosones of parents and recombine to produce
offspring
Evaluate fitness of offspring and repeat mating sequence untila
Ch 6, 7
solution of acceptable fitness evolves
 Genetic Synthesis

“Cut and splice” Uniform

crossover Crossover
Generation of offspring may be allowed to
introduce chromosones of new lengths (cut
and splice) or may be constrained to a
fixed length. Thus the number of
dielectric layers can be held constant or
allowed to be flexible.
 Mirror Reflectivity
Detemrine Mirror reflectivity for each “organism”
by solving eigenvalues of field matirx and
determining reflection coefficient
a0 a1 a2 a3 a4 a6 a7 a8
a5
b0 b1 b2 b3 b4 b6 b7 b8
b5

    
a0 1 0 0 0 . . . a0
 b0   r01 0 0 t01 e−ikn1 x1 . . .  
     b0 
a1  t01 e−ikn1 x1 0 0 0 . . .  
 =  a1 
 b1   0 0 r12 0 . . .  
     b1 
.. .. .. .. .. .. ..
. . . . . . .
Eigenvectors of matrix give steady state field
amplitudes. Ratio of |bo/ao|2 is mirror reflectivity.
 Tantala Thickness
Thickness of Tantala is simply the sum of all odd
genes (assuming we define the odd genes to be
the high index layers)
!
dT a2 O5 = xi
x=1,3,5,...
 Fitness Criteria
A (largely arbitrary) function defining how well
the mirror meets the design goals, i.e.
! "2 ! "2
1−R dT a2 O5
f (R, dT a2 O5 ) = +
15 ppm 5000 nm

We seek to minimize this function, with an

acceptable solution having a value below 1, i.e. a
transmission of less than 15 ppm and a total
tantala thickness of less than 5000 nm.
Genetic Algorithm Pseudo Code

1.! Choose initial population with each gene of each

! organism randomized to a value from 0 to λ/2
2.! Evaluate the fitness f of each organism in the
! population
3.! Repeat until organism with f<1 is found
1.! Select best-ranking individuals to reproduce
2.! Breed new generation through crossover and
! mutation (genetic operations) and give birth to
! offspring
3.! Evaluate the individual fitnesses f of the offspring
4.! Replace worst ranked part of population with
! offspring
Comparing mirrors (15ppm loss)
non-periodic periodic λ/4+λ/4
(44 layers, 7033 nm) (38 layers, 6153nm)
1816 nm Ta2O5 2490 nm Ta2O5
5217 nm SiO2 3663 nm SiO2
15 ppm 15 ppm

sn.
Comparing mirrors (44 layers)
non-periodic periodic 3λ/8+λ/8
(44 layers, 7033 nm) (44 layers, 7766 nm)
1816 nm Ta2O5, 5217 nm SiO2! ! !! ! 1430 nm Ta2O5, 6336 nm SiO2! ! !

15 60

sn.
Periodic coatings 3λ/8 SiO2 +λ/8 Ta2O5

1-R2 sn.15
Comparing mirrors (»15 ppm)

non-periodic periodic (3λ/8, λ/8)

(44 layers, 7033 nm) (52 layers, 9178 nm)
1816 nm Ta2O5 1690 nm Ta2O5
5217 nm SiO2 7488 nm SiO2
15 ppm!! ! 13 ppm!! !

sn.
Non-periodic Coating structure properties

sn.
Number of layers

First attempt
Non-periodic

1-R2 (ppm)
Genetically optimized coating robustness

44 layer, 15ppm design

Introduce 1 nm r.m.s. error on coating

thickness Probability
1.75
density
10,000 trials 1.5

1.25

1
0.55 ppm FWHM
0.75

0.5

0.25

14.25 14.5 14.75 15.25 15.5 15.75

14.25 14.50 14.75 15.00 15.25 15.75 transmittance [ppm]
sn.
Genetically optimized coating bandwidth
1-R2 (ppm)

sn.
 Summary
Non-periodic coatings can be designed by
genetic algorithms;

Multiple heterogeneous (e.g., transmittance,

thickness of constituents, etc.) constraints can
be introduced;

Non λ/4 periodic structures can also be

employed

sn.
 References
Juri Agresti , Giuseppe Castaldi , Riccardo
DeSalvo , Vincenzo Galdi , Vincenzo Pierro c, and
Innocenzo M. Pinto, “Optimized multilayer
dielectric mirror coatings for gravitational wave
interferometers”, LIGO-P060027-00-Z