Probing the atomic structure of mirror coatings
using transmission electron microscopy
Stuart Reid, Riccardo Bassiri1 , Konstantin B. Borisenko2 , David J. H.
Cockayne2 , Keith Evans1, James Hough1, Ian MacLaren1, Iain Martin1,
Sheila Rowan1
1
SUPA, University of Glasgow, 2 Department of Materials, University of Oxford
GWADW, Kyoto, Japan – May 2010
1
Coating development critical for the future
-19
10
-20
10
1st generation
10
Strain [Hz
-1/2
]
-21
2nd generation
-22
10
3rd generation
-23
10
-24
10
-25
10
1
10
100
Frequency [Hz]
2
Coating thermal
Noise
1000
10000
Introduction
• The test mass mirror coatings are estimated to be a significant source of
thermal noise in future ground-based GW detectors
• Thermal noise is proportional to the mechanical loss
(internal friction) of the material
GEO600 test-mass
• Considerable research is being conducted into understanding the material
properties of these coatings (see previous and following talks)
– The focus here is on the atomic structure and how this affects the
material properties - and in particular the mechanical loss
What is causing the mechanical loss on
an atomic level?
3
Transmission electron microscopy
• Useful for probing atomic structure and chemistry
Electron
beam
Sample
Direct beam
Transmission Electron Microscope
Tecnai T-20
Interactions of electron beam
with sample
• Allows us to characterise atomic structure
– Imaging
– Diffraction
– Spectroscopy
4
Transmission electron microscopy
Initial interesting results:
Image of multilayer coating,
(bright- silica, dark - tantala)
(from Ta2O5 samples heat-treated at a range of temperatures)
Amorphous diffraction pattern
of 300oC tantala
Crystalline diffraction pattern
of 800oC tantala
• Compare TEM results to mechanical loss
measurements
• The 800oC sample has high loss peak at 8090K probably due to crystallisation
• To probe the properties of the amorphous
samples we need Reduced Radial Density
Functions
5
Mechanical loss measurements for heat treated tantala (see previous talk by Matt Abernathy
Reduced radial density functions
•
•
Silica and tantala are amorphous materials
– They do not have long range order
– They do have short range order
We can probe this short range order with reduced density functions
– RDFs give a statistical representation of where atoms sit with regards to a central
atom
Tantala diffraction pattern
Intensity profile
Reduced density function
• The reduced density function is a Fourier transform of the information
gained from the intensity profile [D. J. H. Cockayne, Annu.Rev.Mater.Res, 37:159-87, (2007)]
6
Reduced density functions
• Three Ta2O5 coatings were measured
– Each one was heat-treated at a different temperature (300, 400 & 600oC)
RDFs of heat-treated Ta2O5
– RDFs show differences in local atomic structure as heat treatment temperature rises
– From comparison to the structure of crystalline Ta2O5 we can deduce that the first peak
arises from Ta - O bonds and second peak from Ta - Ta bonds
– Both first and second peaks become more defined and difference in heights between
them decrease as heat treatment temperature rises implying that:
– Material is becoming more ordered
7
– There is an increase in Ta - Ta bonding
Modelling the atomic structure
•
Why?
– If we accurately interpret the RDF
– What do the peaks mean?
– What bond types correspond to each peak?
•
We can then:
– Investigate the atomic structure
Reduced density function
– Average distances between atoms
– Co-ordination numbers
– Bond types
– Bond angles
– Probe the material properties
– Relationship to mechanical loss
– Optical properties?
– Other material properties?
8
Energy optimised Ta2O5 atomic model
(blue - Ta, red - O)
Modelling the atomic structure
•
Reverse Monte Carlo refinements comparing model to experimental RDFs
RMC refinement
Energy optimisation
•
Constrains the model
further
•
Ensure atoms are sitting in
a physically stable position
•
9
Gives a greater degree of
accuracy
Do experimental and
theoretical RDFs agree?
Constraints
Initial constraints from the
crystalline structure of Ta2O5
– Atom types
Final structure
– Bond lengths
– Bond types
– Bond distributions
Modelling the atomic structure
RDF comparison before RMC (using initial boundary conditions model)
Refinement process
RDF comparison after RMC (using RMC + energy optimised model)
10
Modelling the atomic structure
• Reverse Monte Carlo modelling was carried out on the 400oC heat
treated tantala coating
– Results from this model show an
average Ta to Ta bond length of 3.19Å
and Ta to O bond length of 1.99Å
– Co-ordination number for Ta = 6.53,
O=2.09
Crystalline model of Ta2O5
(Aleshina et al., Cryst. Rep. 47, 2002)
– Ring structure of Ta and O bonds
remains partially intact from the
crystalline phase
Energy optimised Ta2O5 atomic model
(Blue - Ta ,Red - O)
11
Modelling the atomic structure
• Partial RDF data:
– Allows a greater understanding
of the relative distances from
one atom to another
– Shows exactly what the peaks in
the initial RDF mean
– Initial assumptions on comparing
the peaks to the crystal phase
are accurate
RDFs of heat-treated Ta2O5
– Ta - O bonds dominate first
peak
– Ta - Ta bonds dominate second
peak
Partial RDF of 400oC model
12
Modelling the atomic structure
• Bond types and angles:
– Atomic modelling makes
understanding bond structures
in the sample easier
– Bond types
– Bond angle distributions
Bond type distribution
Bond angle type distribution
• Provides an excellent way to
compare the changes in the
atomic structure
13
Bond angle distribution of 400oC model
Future work
• Near future
– Compare results from each of
the three heat-treated atomic
models
– Start modelling Ti doped and
low water Ta2O5 samples
(similar process)
• Future
– Investigate ways of getting
material properties from
models
– X-ray scattering measurements
for single element ‘RDF’
analysis of Ti doped samples
14
RDFs of heat-treated Ta2O5
Ti doped tantala RDFs
Conclusion
What is causing the mechanical loss on an atomic level?
• Significant progress towards answering this question
-
Now have well developed techniques in order to probe the atomic structure
• For Ta2O5 coatings heat-treated to 300, 400, 600 and 800oC:
-
Samples heat-treated to 300, 400, 600oC are amorphous, the 800oC sample has
crystallised possibly causing the high mechanical loss peak at low temperatures
-
Preliminary results from the 300, 400 and 600oC show an increase in local ordering and
number of Ta - Ta bonds as heat treatment temperature increases
-
Atomic modelling provides an accurate way to fully understand the RDF and investigate
bond types and distributions
• Combining microscopic techniques together with mechanical loss
measurements will allow us to gain a better understanding of how these
mirror coatings perform and help produce low mechanical loss coatings
• The same techniques will be applied to other mirror coatings that have
varying material properties
15